Index: Orange/OrangeWidgets/OWNxHist.py =================================================================== --- Orange/OrangeWidgets/OWNxHist.py (revision 10516) +++ Orange/OrangeWidgets/OWNxHist.py (revision 10499) @@ -30,9 +30,6 @@ self.andor = 0 self.matrix = None - self.excludeLimit = 2 + self.excludeLimit = 1 self.percentil = 0 - - self.graph = None - self.graph_matrix = None if parent is None: @@ -61,5 +58,5 @@ hb = OWGUI.widgetBox(ribg, None, orientation="horizontal", addSpace=False) OWGUI.appendRadioButton(ribg, self, "netOption", "Large components only. Min nodes:", insertInto=hb, callback=self.generateGraph) - OWGUI.spin(hb, self, "excludeLimit", 2, 100, 1, callback=(lambda h=True: self.generateGraph(h))) + OWGUI.spin(hb, self, "excludeLimit", 1, 100, 1, callback=(lambda h=True: self.generateGraph(h))) OWGUI.appendRadioButton(ribg, self, "netOption", "Largest connected component only", callback=self.generateGraph) OWGUI.appendRadioButton(ribg, self, "netOption", "Connected component with vertex") @@ -81,7 +78,4 @@ def setPercentil(self): - if self.matrix is None or self.percentil <= 0: - return - self.spinLowerThreshold = self.histogram.minValue # flatten matrix, sort values and remove identities (self.matrix[i][i]) @@ -100,7 +94,4 @@ def setMatrix(self, data): if data == None: return - - if not hasattr(data, "items") or data.items is None: - setattr(data, "items", [i for i in range(data.dim)]) self.matrix = data @@ -114,5 +105,5 @@ upp = max(values) self.spinLowerThreshold = self.spinUpperThreshold = math.floor(low - (0.03 * (upp - low))) - + self.generateGraph() self.attributeCombo.clear() vars = [] @@ -134,7 +125,4 @@ except: print "error adding ", var, " to the attribute combo" - - self.setPercentil() - self.generateGraph() def changeLowerSpin(self): @@ -192,5 +180,5 @@ matrix = self.matrix - if hasattr(self.matrix, "items") and self.matrix.items is not None: + if hasattr(self.matrix, "items"): if type(self.matrix.items) == Orange.data.Table: graph.set_items(self.matrix.items) @@ -214,5 +202,5 @@ # exclude unconnected if str(self.netOption) == '1': - components = [x for x in Orange.network.nx.algorithms.components.connected_components(graph) if len(x) >= self.excludeLimit] + components = [x for x in Orange.network.nx.algorithms.components.connected_components(graph) if len(x) > self.excludeLimit] if len(components) > 0: include = reduce(lambda x, y: x + y, components) @@ -230,5 +218,5 @@ component = Orange.network.nx.algorithms.components.connected_components(graph)[0] if len(component) > 1: - self.graph = graph.subgraph(component) + self.graph = graph.subgraph(include) matrix = self.matrix.getitems(component) else: Index: Orange/OrangeWidgets/Prototypes/OWBR.py =================================================================== --- Orange/OrangeWidgets/Prototypes/OWBR.py (revision 10502) +++ Orange/OrangeWidgets/Prototypes/OWBR.py (revision 9671) @@ -59,6 +59,5 @@ if not Orange.multilabel.is_multilabel(data): - self.warning(0, "Multi-label data with class values 0 and 1 is " - "expected on the input.") + self.warning(0, "Multi-label data is expected on the input.") return self.warning(0, None) Index: Orange/OrangeWidgets/Prototypes/OWBRkNN.py =================================================================== --- Orange/OrangeWidgets/Prototypes/OWBRkNN.py (revision 10502) +++ Orange/OrangeWidgets/Prototypes/OWBRkNN.py (revision 9671) @@ -63,6 +63,5 @@ if not Orange.multilabel.is_multilabel(data): - self.warning(0, "Multi-label data with class values 0 and 1 is " - "expected on the input.") + self.warning(0, "Multi-label data is expected on the input.") return self.warning(0, None) Index: range/OrangeWidgets/Prototypes/OWCSVFileImport.py =================================================================== --- Orange/OrangeWidgets/Prototypes/OWCSVFileImport.py (revision 10529) +++ (revision ) @@ -1,310 +1,0 @@ -""" -CSV File import -Import .csv file - -""" -import os -import csv -from StringIO import StringIO - -from PyQt4 import QtCore, QtGui - -from OWWidget import * -import OWGUI -import Orange - -#from OWFile import FileNameContextHandler as PathContextHandler -from OWDataTable import ExampleTableModel - -DEFAULT_OPTIONS = {"delimiter":",", - "quote":"'", - "has_header": True, - "has_orange_definitions": False, - } - -Slot = pyqtSlot - -class standard_icons(object): - def __init__(self, qwidget=None, style=None): - self.qwidget = qwidget - if qwidget is None: - self.style = QApplication.instance().style() - else: - self.style = qwidget.style() - - @property - def dir_open_icon(self): - return self.style.standardIcon(QStyle.SP_DirOpenIcon) - - @property - def reload_icon(self): - return self.style.standardIcon(QStyle.SP_BrowserReload) - - - - -class OWCSVFileImport(OWWidget): -# contextHandlers = {"": PathContextHandler("")} - settingsList = ["recent_files"] - - DELIMITERS = [("Tab", "\t"), - ("Comma", ","), - ("Semicolon", ";"), - ("Space", " "), - ("Others", None), - ] - def __init__(self, parent=None, signalManager=None, - title="CSV File Import"): - - OWWidget.__init__(self, parent, signalManager, title, - wantMainArea=False) - - self.inputs = [] - self.outputs = [("Data", Orange.data.Table)] - - # Settings - self.delimiter = "," - self.quote = '"' - - self.has_header = True - self.has_orange_header = True - - self.recent_files = [] - self.hints = {} - - self.loadSettings() - - self.recent_file = filter(os.path.exists, self.recent_files) - - self.csv_options = dict(DEFAULT_OPTIONS) - - layout = QHBoxLayout() - box = OWGUI.widgetBox(self.controlArea, "File", orientation=layout) - self.style().standardIcon(QStyle.SP_DirOpenIcon) - icons = standard_icons(self) - - self.recent_combo = QComboBox(self, objectName="recent_combo", - toolTip="Recent files.", - activated=self.on_select_recent) - self.recent_combo.addItems([os.path.basename(p) for p in self.recent_files]) - - self.browse_button = QPushButton("...", icon=icons.dir_open_icon, - toolTip="Browse filesystem", - clicked=self.on_open_dialog) - -# self.reload_button = QPushButton("Reload", icon=icons.reload_icon, -# clicked=self.reload_file) - - layout.addWidget(self.recent_combo, 2) - layout.addWidget(self.browse_button) -# layout.addWidget(self.reload_button) - - form_left = QFormLayout() - form_right = QFormLayout() - h_layout = QHBoxLayout() - grid_layout = QGridLayout() - grid_layout.setVerticalSpacing(4) - grid_layout.setHorizontalSpacing(4) - box = OWGUI.widgetBox(self.controlArea, "Cell Separator", - orientation=grid_layout) - - button_group = QButtonGroup(box) - QObject.connect(button_group, - SIGNAL("buttonPressed(int)"), - self.delimiter_changed) -# button_group.buttonPressed.connect(self.delimiter_changed) - - for i, (name, char) in enumerate(self.DELIMITERS[:-1]): - button = QRadioButton(name, box, - toolTip="Use %r as cell separator" % char) - - button_group.addButton(button, i) - grid_layout.addWidget(button, i / 3, i % 3) - - button = QRadioButton("Other", box, - toolTip="Use other character") - - button_group.addButton(button, i + 1) - grid_layout.addWidget(button, i / 3 + 1, 0) - self.delimiter_button_group = button_group - - self.delimiter_edit = QLineEdit(objectName="delimiter_edit", - text=self.delimiter, - editingFinished=self.delimiter_changed, - toolTip="Cell delimiter character.") - grid_layout.addWidget(self.delimiter_edit, i / 3 + 1, 1, -1, -1) - - self.quote_edit = QLineEdit(objectName="quote_edit", - text=self.quote, - editingFinished=self.quote_changed, - toolTip="Text quote character.") - - form = QFormLayout() - box = OWGUI.widgetBox(self.controlArea, "Other Options", - orientation=form) - -# form_left.addRow("Delimiter", self.delimiter_edit) - form.addRow("Quote", self.quote_edit) - - self.has_header_check = \ - QCheckBox(objectName="has_header_check", - checked=self.has_header, - text="Header line", - toolTip="Use the first line as a header", - toggled=self.has_header_changed - ) - - form.addRow(self.has_header_check) - - self.has_orange_header_check = \ - QCheckBox(objectName="has_orange_header_check", - checked=self.has_orange_header, - text="Has orange variable type definitions", - toolTip="Use second and third line as a orange style '.tab' format feature definitions.", - toggled=self.has_orange_header_changed - ) - - form.addRow(self.has_orange_header_check) - - box = OWGUI.widgetBox(self.controlArea, "Preview") - self.preview_view = QTableView() - box.layout().addWidget(self.preview_view) - - OWGUI.button(self.controlArea, self, "Send", callback=self.send_data) - - self.selected_file = None - self.data = None - - self.resize(400, 350) - - def on_select_recent(self, recent): - if isinstance(recent, int): - recent = self.recent_files[recent] - - self.set_selected_file(recent) - - def on_open_dialog(self): - last = os.path.expanduser("~/Documents") - path = QFileDialog.getOpenFileName(self, "Open File", last) - path = unicode(path) - if path: - basedir, name = os.path.split(path) - self.recent_combo.insertItem(0, name) - self.recent_files.insert(0, path) - self.set_selected_file(path) - - @Slot(int) - def delimiter_changed(self, index): - self.delimiter = self.DELIMITERS[index][1] - if self.delimiter is None: - self.other_delimiter = str(self.delimiter_edit.text()) - self.update_preview() - - def quote_changed(self): - if self.quote_edit.text(): - self.quote = str(self.quote_edit.text()) - self.update_preview() - - def has_header_changed(self): - self.has_header = self.has_header_check.isChecked() - self.update_preview() - - def has_orange_header_changed(self): - self.has_orange_header = self.has_orange_header_check.isChecked() - self.update_preview() - - def set_selected_file(self, filename): - if filename in self.hints: - hints = self.hints[filename] - else: - hints = sniff_csv(filename) - self.hints[filename] = hints - - delimiter = hints["delimiter"] - preset = [d[1] for d in self.DELIMITERS[:-1]] - if delimiter not in preset: - self.delimiter = None - self.other_delimiter = delimiter - self.delimiter_edit.setText(self.other_delimiter) - self.delimiter_edit.setEnabled(True) - else: - index = preset.index(delimiter) - button = self.delimiter_button_group.button(index) - button.setChecked(True) - self.delimiter_edit.setEnabled(False) - - self.quote = hints["quotechar"] - self.quote_edit.setText(self.quote) - - self.has_header = hints["has_header"] - self.has_header_check.setChecked(self.has_header) - - self.has_orange_header = hints["has_orange_header"] - self.has_orange_header_check.setChecked(self.has_orange_header) - - self.selected_file = filename - self.selected_file_head = [] - with open(self.selected_file, "rb") as f: - for i, line in zip(range(30), f): - self.selected_file_head.append(line) - - self.update_preview() - - def update_preview(self): - if self.selected_file: - head = StringIO("".join(self.selected_file_head)) - hints = self.hints[self.selected_file] - - hints["quote"] = self.quote - hints["delimiter"] = self.delimiter or self.other_delimiter - hints["has_header"] = self.has_header - hints["has_orange_header"] = self.has_orange_header - - try: - data = Orange.data.io.load_csv(head, delimiter=self.delimiter, - quotechar=self.quote, - has_header=self.has_header, - has_types=self.has_orange_header, - has_annotations=self.has_orange_header, - ) - except Exception, ex: - self.error(0, "Cannot parse (%r)" % ex) - self.data = None - return - model = ExampleTableModel(data, None, self) - self.preview_view.setModel(model) - self.data = data - - def send_data(self): - self.send("Data", self.data) - - -def sniff_csv(file): - snifer = csv.Sniffer() - if isinstance(file, basestring): - file = open(file, "rb") - - sample = file.read(2 ** 20) # max 1MB sample - dialect = snifer.sniff(sample) - has_header = snifer.has_header(sample) - - return {"delimiter": dialect.delimiter, - "doublequote": dialect.doublequote, - "escapechar": dialect.escapechar, - "quotechar": dialect.quotechar, - "quoting": dialect.quoting, - "skipinitialspace": dialect.skipinitialspace, - "has_header": has_header, - "has_orange_header": False} - -if __name__ == "__main__": - import sys - app = QApplication(sys.argv) - w = OWCSVFileImport() - w.show() - app.exec_() - w.saveSettings() - - - - Index: Orange/OrangeWidgets/Prototypes/OWLP.py =================================================================== --- Orange/OrangeWidgets/Prototypes/OWLP.py (revision 10502) +++ Orange/OrangeWidgets/Prototypes/OWLP.py (revision 9671) @@ -59,6 +59,5 @@ if not Orange.multilabel.is_multilabel(data): - self.warning(0, "Multi-label data with class values 0 and 1 is " - "expected on the input.") + self.warning(0, "Multi-label data is expected on the input.") return self.warning(0, None) Index: Orange/OrangeWidgets/Prototypes/OWMLkNN.py =================================================================== --- Orange/OrangeWidgets/Prototypes/OWMLkNN.py (revision 10502) +++ Orange/OrangeWidgets/Prototypes/OWMLkNN.py (revision 9671) @@ -72,6 +72,5 @@ if not Orange.multilabel.is_multilabel(data): - self.warning(0, "Multi-label data with class values 0 and 1 is " - "expected on the input.") + self.warning(0, "Multi-label data is expected on the input.") return self.warning(0, None) Index: Orange/OrangeWidgets/Unsupervised/OWNetClustering.py =================================================================== --- Orange/OrangeWidgets/Unsupervised/OWNetClustering.py (revision 9671) +++ Orange/OrangeWidgets/Unsupervised/OWNetClustering.py (revision 9671) @@ -0,0 +1,49 @@ +""" +THIS WIDGET IS OBSOLETE; USE OWNxClustering.py +""" + +import orange +import orngNetwork +import OWGUI + +from OWWidget import * + +class OWNetClustering(OWWidget): + + settingsList = ['method', 'iterationHistory', "autoApply"] + + def __init__(self, parent=None, signalManager=None): + OWWidget.__init__(self, parent, signalManager, 'Network Clustering') + + self.inputs = [("Network", orngNetwork.Network, self.setNetwork, Default)] + self.outputs = [("Network", orngNetwork.Network)] + + self.net = None + self.method = 0 + self.iterationHistory = 0 + self.autoApply = 0 + + self.loadSettings() + + ribg = OWGUI.radioButtonsInBox(self.controlArea, self, "method", [], "Method", callback = self.cluster) + OWGUI.appendRadioButton(ribg, self, "method", "Label propagation clustering (Raghavan et al., 2007)", callback = self.cluster) + OWGUI.checkBox(OWGUI.indentedBox(ribg), self, "iterationHistory", "Append clustering data on each iteration", callback = self.cluster) + self.info = OWGUI.widgetLabel(self.controlArea, ' ') + autoApplyCB = OWGUI.checkBox(self.controlArea, self, "autoApply", "Commit automatically") + OWGUI.button(self.controlArea, self, "Commit", callback=self.cluster) + + def setNetwork(self, net): + self.net = net + if self.autoApply: + self.cluster() + + def cluster(self): + self.info.setText(' ') + + if self.net == None: + self.send("Network", None) + return + + labels = self.net.clustering.labelPropagation(results2items=1, resultHistory2items=self.iterationHistory) + self.info.setText('%d clusters found' % len(set(labels))) + self.send("Network", self.net) Index: Orange/OrangeWidgets/Unsupervised/OWNetExplorer.py =================================================================== --- Orange/OrangeWidgets/Unsupervised/OWNetExplorer.py (revision 10466) +++ Orange/OrangeWidgets/Unsupervised/OWNetExplorer.py (revision 10466) @@ -0,0 +1,1791 @@ +""" +THIS WIDGET IS OBSOLETE; USE OWNxExplorer.py +""" +import orange +import OWGUI +import OWColorPalette +import orngNetwork +import OWToolbars +import math +import operator +import orngMDS + +from OWWidget import * +from OWNetworkCanvas import * +from time import * +from statc import mean +from operator import itemgetter + +dir = os.path.dirname(__file__) + "/../icons/" +dlg_mark2sel = dir + "Dlg_Mark2Sel.png" +dlg_sel2mark = dir + "Dlg_Sel2Mark.png" +dlg_selIsmark = dir + "Dlg_SelisMark.png" +dlg_selected = dir + "Dlg_SelectedNodes.png" +dlg_unselected = dir + "Dlg_UnselectedNodes.png" +dlg_showall = dir + "Dlg_clear.png" + +class OWNetExplorer(OWWidget): + settingsList = ["autoSendSelection", "spinExplicit", "spinPercentage", + "maxLinkSize", "minVertexSize", "maxVertexSize", "renderAntialiased", + "invertSize", "optMethod", "lastVertexSizeColumn", "lastColorColumn", + "lastNameComponentAttribute", "lastLabelColumns", "lastTooltipColumns", + "showWeights", "showIndexes", "showEdgeLabels", "colorSettings", + "selectedSchemaIndex", "edgeColorSettings", "selectedEdgeSchemaIndex", + "showMissingValues", "fontSize", "mdsTorgerson", "mdsAvgLinkage", + "mdsSteps", "mdsRefresh", "mdsStressDelta", "organism","showTextMiningInfo", + "toolbarSelection", "minComponentEdgeWidth", "maxComponentEdgeWidth", + "mdsFromCurrentPos", "labelsOnMarkedOnly", "tabIndex"] + + def __init__(self, parent=None, signalManager=None, name = 'Net Explorer', + NetworkCanvas=OWNetworkCanvas): + OWWidget.__init__(self, parent, signalManager, name) + #self.contextHandlers = {"": DomainContextHandler("", [ContextField("attributes", selected="markerAttributes"), ContextField("attributes", selected="tooltipAttributes"), "color"])} + self.inputs = [("Network", orngNetwork.Network, self.setGraph, Default), + ("Items", orange.ExampleTable, self.setItems), + ("Marked Items", orange.ExampleTable, self.markItems), + ("Item Subset", orange.ExampleTable, self.setExampleSubset), + ("Vertex Distance", orange.SymMatrix, self.setVertexDistance)] + + self.outputs = [("Selected Network", orngNetwork.Network), + ("Selected Distance Matrix", orange.SymMatrix), + ("Selected Data", ExampleTable), + ("Other Data", ExampleTable), + ("Marked Data", ExampleTable), + ("Features", AttributeList)] + + self.markerAttributes = [] + self.tooltipAttributes = [] + self.edgeLabelAttributes = [] + self.attributes = [] + self.edgeAttributes = [] + self.autoSendSelection = False + self.graphShowGrid = 1 # show gridlines in the graph + + self.markNConnections = 2 + self.markNumber = 0 + self.markProportion = 0 + self.markSearchString = "" + self.markDistance = 2 + self.frSteps = 1 + self.hubs = 0 + self.color = 0 + self.edgeColor = 0 + self.vertexSize = 0 + self.nVertices = self.nShown = self.nHidden = self.nMarked = self.nSelected = self.nEdges = self.verticesPerEdge = self.edgesPerVertex = self.diameter = self.clustering_coefficient = 0 + self.optimizeWhat = 1 + self.stopOptimization = 0 + self.maxLinkSize = 3 + self.maxVertexSize = 5 + self.minVertexSize = 5 + self.renderAntialiased = 1 + self.labelsOnMarkedOnly = 0 + self.invertSize = 0 + self.optMethod = 0 + self.lastVertexSizeColumn = '' + self.lastColorColumn = '' + self.lastNameComponentAttribute = '' + self.lastLabelColumns = set() + self.lastTooltipColumns = set() + self.showWeights = 0 + self.showIndexes = 0 + self.showEdgeLabels = 0 + self.colorSettings = None + self.selectedSchemaIndex = 0 + self.edgeColorSettings = [('net_edges', [[], [('contPalette', (4294967295L, 4278190080L, 0))], [('discPalette', [(204, 204, 204), (179, 226, 205), (253, 205, 172), (203, 213, 232), (244, 202, 228), (230, 245, 201), (255, 242, 174), (241, 226, 204)])]]), ('Default', [[], [('contPalette', (4294967295L, 4278190080L, 0))], [('discPalette', [(0, 0, 255), (255, 0, 0), (0, 255, 0), (255, 128, 0), (255, 255, 0), (255, 0, 255), (0, 255, 255), (128, 0, 255), (0, 128, 255), (255, 223, 128), (127, 111, 64), (92, 46, 0), (0, 84, 0), (192, 192, 0), (0, 127, 127), (128, 0, 0), (127, 0, 127)])]])] + self.selectedEdgeSchemaIndex = 0 + self.vertexDistance = None + self.showDistances = 0 + self.showMissingValues = 0 + self.fontSize = 12 + self.mdsTorgerson = 0 + self.mdsAvgLinkage = 0 + self.mdsSteps = 120 + self.mdsRefresh = 30 + self.mdsStressDelta = 0.00001 + self.organism = 'goa_human' + self.showTextMiningInfo = 0 + self.toolbarSelection = 0 + self.minComponentEdgeWidth = 0 + self.maxComponentEdgeWidth = 0 + self.mdsFromCurrentPos = 0 + self.tabIndex = 0 + + self.loadSettings() + + self.optimization = None + self.markInputItems = None + + self.mainArea.layout().setContentsMargins(0,4,4,4) + self.controlArea.layout().setContentsMargins(4,4,0,4) + + self.networkCanvas = NetworkCanvas(self, self.mainArea, "Net Explorer") + self.networkCanvas.showMissingValues = self.showMissingValues + self.mainArea.layout().addWidget(self.networkCanvas) + + self.networkCanvas.maxLinkSize = self.maxLinkSize + self.networkCanvas.minVertexSize = self.minVertexSize + self.networkCanvas.maxVertexSize = self.maxVertexSize + + self.hcontroArea = OWGUI.widgetBox(self.controlArea, orientation='horizontal') + + self.tabs = OWGUI.tabWidget(self.hcontroArea) + self.connect(self.tabs, SIGNAL("currentChanged(int)"), self.currentTabChanged) + + self.verticesTab = OWGUI.createTabPage(self.tabs, "Vertices") + self.edgesTab = OWGUI.createTabPage(self.tabs, "Edges") + self.markTab = OWGUI.createTabPage(self.tabs, "Mark") + self.infoTab = OWGUI.createTabPage(self.tabs, "Info") + #self.editTab = OWGUI.createTabPage(self.tabs, "Edit") + + + self.optimizeBox = OWGUI.radioButtonsInBox(self.verticesTab, self, "optimizeWhat", [], "Optimize", addSpace=False) + + self.optCombo = OWGUI.comboBox(self.optimizeBox, self, "optMethod", label='Method: ', orientation='horizontal', callback=self.setOptMethod) + self.optCombo.addItem("No optimization") + self.optCombo.addItem("Random") + self.optCombo.addItem("Fruchterman Reingold") + self.optCombo.addItem("Fruchterman Reingold Weighted") + self.optCombo.addItem("Fruchterman Reingold Radial") + self.optCombo.addItem("Circular Crossing Reduction") + self.optCombo.addItem("Circular Original") + self.optCombo.addItem("Circular Random") + self.optCombo.addItem("Pivot MDS") + self.optCombo.setCurrentIndex(self.optMethod) + self.stepsSpin = OWGUI.spin(self.optimizeBox, self, "frSteps", 1, 100000, 1, label="Iterations: ") + self.stepsSpin.setEnabled(False) + + self.optButton = OWGUI.button(self.optimizeBox, self, "Optimize layout", callback=self.optLayout, toggleButton=1) + + colorBox = OWGUI.widgetBox(self.verticesTab, "Vertex color attribute", orientation="horizontal", addSpace = False) + self.colorCombo = OWGUI.comboBox(colorBox, self, "color", callback=self.setVertexColor) + self.colorCombo.addItem("(same color)") + OWGUI.button(colorBox, self, "Set vertex color palette", self.setColors, tooltip = "Set vertex color palette", debuggingEnabled = 0) + + self.vertexSizeCombo = OWGUI.comboBox(self.verticesTab, self, "vertexSize", box = "Vertex size attribute", callback=self.setVertexSize) + self.vertexSizeCombo.addItem("(none)") + + OWGUI.spin(self.vertexSizeCombo.box, self, "minVertexSize", 5, 200, 1, label="Min vertex size:", callback = self.setVertexSize) + OWGUI.spin(self.vertexSizeCombo.box, self, "maxVertexSize", 5, 200, 1, label="Max vertex size:", callback = self.setVertexSize) + OWGUI.checkBox(self.vertexSizeCombo.box, self, "invertSize", "Invert vertex size", callback = self.setVertexSize) + + colorBox = OWGUI.widgetBox(self.edgesTab, "Edge color attribute", orientation="horizontal", addSpace = False) + self.edgeColorCombo = OWGUI.comboBox(colorBox, self, "edgeColor", callback=self.setEdgeColor) + self.edgeColorCombo.addItem("(same color)") + OWGUI.button(colorBox, self, "Set edge color palette", self.setEdgeColorPalette, tooltip = "Set edge color palette", debuggingEnabled = 0) + + self.attBox = OWGUI.widgetBox(self.verticesTab, "Vertex labels | tooltips", orientation="vertical", addSpace = False) + OWGUI.spin(self.attBox, self, "fontSize", 4, 30, 1, label="Set font size:", callback = self.setFontSize) + + self.attBox = OWGUI.widgetBox(self.attBox, orientation="horizontal", addSpace = False) + self.attListBox = OWGUI.listBox(self.attBox, self, "markerAttributes", "attributes", selectionMode=QListWidget.MultiSelection, callback=self.clickedAttLstBox) + self.tooltipListBox = OWGUI.listBox(self.attBox, self, "tooltipAttributes", "attributes", selectionMode=QListWidget.MultiSelection, callback=self.clickedTooltipLstBox) + + self.edgeLabelBox = OWGUI.widgetBox(self.edgesTab, "Edge labels", addSpace = False) + self.edgeLabelListBox = OWGUI.listBox(self.edgeLabelBox, self, "edgeLabelAttributes", "edgeAttributes", selectionMode=QListWidget.MultiSelection, callback=self.clickedEdgeLabelListBox) + self.edgeLabelBox.setEnabled(False) + + ib = OWGUI.widgetBox(self.edgesTab, "General", orientation="vertical") + OWGUI.checkBox(ib, self, 'showWeights', 'Show weights', callback = self.showWeightLabels) + OWGUI.checkBox(ib, self, 'showEdgeLabels', 'Show labels on edges', callback = self.showEdgeLabelsClick) + OWGUI.spin(ib, self, "maxLinkSize", 1, 50, 1, label="Max edge width:", callback = self.setMaxLinkSize) + OWGUI.checkBox(ib, self, 'showDistances', 'Explore vertex distances', callback = self.showDistancesClick) + + ib = OWGUI.widgetBox(self.verticesTab, "General", orientation="vertical") + OWGUI.checkBox(ib, self, 'showIndexes', 'Show indexes', callback = self.showIndexLabels) + OWGUI.checkBox(ib, self, 'labelsOnMarkedOnly', 'Show labels on marked vertices only', callback = self.labelsOnMarked) + OWGUI.checkBox(ib, self, 'renderAntialiased', 'Render antialiased', callback = self.setRenderAntialiased) + self.insideView = 0 + self.insideViewNeighbours = 2 + OWGUI.spin(ib, self, "insideViewNeighbours", 1, 6, 1, label="Inside view (neighbours): ", checked = "insideView", checkCallback = self.insideview, callback = self.insideviewneighbours) + OWGUI.checkBox(ib, self, 'showMissingValues', 'Show missing values', callback = self.setShowMissingValues) + + ib = OWGUI.widgetBox(self.markTab, "Info", orientation="vertical") + OWGUI.label(ib, self, "Vertices (shown/hidden): %(nVertices)i (%(nShown)i/%(nHidden)i)") + OWGUI.label(ib, self, "Selected and marked vertices: %(nSelected)i - %(nMarked)i") + + ribg = OWGUI.radioButtonsInBox(self.markTab, self, "hubs", [], "Method", callback = self.setMarkMode) + OWGUI.appendRadioButton(ribg, self, "hubs", "None", callback = self.setMarkMode) + OWGUI.appendRadioButton(ribg, self, "hubs", "Find vertices", callback = self.setMarkMode) + self.ctrlMarkSearchString = OWGUI.lineEdit(OWGUI.indentedBox(ribg), self, "markSearchString", callback=self.setSearchStringTimer, callbackOnType=True) + self.searchStringTimer = QTimer(self) + self.connect(self.searchStringTimer, SIGNAL("timeout()"), self.setMarkMode) + + OWGUI.appendRadioButton(ribg, self, "hubs", "Mark neighbours of focused vertices", callback = self.setMarkMode) + OWGUI.appendRadioButton(ribg, self, "hubs", "Mark neighbours of selected vertices", callback = self.setMarkMode) + ib = OWGUI.indentedBox(ribg, orientation = 0) + self.ctrlMarkDistance = OWGUI.spin(ib, self, "markDistance", 0, 100, 1, label="Distance ", callback=(lambda h=2: self.setMarkMode(3 if self.hubs==3 else h))) + #self.ctrlMarkFreeze = OWGUI.button(ib, self, "&Freeze", value="graph.freezeNeighbours", toggleButton = True) + OWGUI.widgetLabel(ribg, "Mark vertices with ...") + OWGUI.appendRadioButton(ribg, self, "hubs", "at least N connections", callback = self.setMarkMode) + OWGUI.appendRadioButton(ribg, self, "hubs", "at most N connections", callback = self.setMarkMode) + self.ctrlMarkNConnections = OWGUI.spin(OWGUI.indentedBox(ribg), self, "markNConnections", 0, 1000000, 1, label="N ", callback=(lambda h=4: self.setMarkMode(h))) + OWGUI.appendRadioButton(ribg, self, "hubs", "more connections than any neighbour", callback = self.setMarkMode) + OWGUI.appendRadioButton(ribg, self, "hubs", "more connections than avg neighbour", callback = self.setMarkMode) + OWGUI.appendRadioButton(ribg, self, "hubs", "most connections", callback = self.setMarkMode) + ib = OWGUI.indentedBox(ribg) + self.ctrlMarkNumber = OWGUI.spin(ib, self, "markNumber", 0, 1000000, 1, label="Number of vertices" + ": ", callback=(lambda h=8: self.setMarkMode(h))) + OWGUI.widgetLabel(ib, "(More vertices are marked in case of ties)") + self.markInputRadioButton = OWGUI.appendRadioButton(ribg, self, "hubs", "Mark vertices given in the input signal", callback = self.setMarkMode) + ib = OWGUI.indentedBox(ribg) + self.markInput = 0 + self.markInputCombo = OWGUI.comboBox(ib, self, "markInput", callback=(lambda h=9: self.setMarkMode(h))) + self.markInputRadioButton.setEnabled(False) + + ib = OWGUI.widgetBox(self.markTab, "General", orientation="vertical") + self.checkSendMarkedNodes = 0 + OWGUI.checkBox(ib, self, 'checkSendMarkedNodes', 'Send marked vertices', callback = self.setSendMarkedNodes, disabled=0) + + + T = OWToolbars.NavigateSelectToolbar + self.zoomSelectToolbar = T(self, self.hcontroArea, self.networkCanvas, self.autoSendSelection, + buttons = (T.IconZoom, + T.IconZoomExtent, + T.IconZoomSelection, + T.IconPan, + ("", "", "", None, None, 0), + #("Move selection", "buttonMoveSelection", "activateMoveSelection", QIcon(OWToolbars.dlg_select), Qt.ArrowCursor, 1), + T.IconRectangle, + #T.IconPolygon, + T.IconSendSelection, + ("", "", "", None, None, 0, "select"), + ("Add marked to selection", "buttonM2S", "markedToSelection", QIcon(dlg_mark2sel), Qt.ArrowCursor, 0), + ("Add selection to marked", "buttonS2M", "selectionToMarked", QIcon(dlg_sel2mark), Qt.ArrowCursor, 0), + ("Remove selection", "buttonRMS", "removeSelection", QIcon(dlg_selIsmark), Qt.ArrowCursor, 0), + ("", "", "", None, None, 0, "select"), + ("Hide selected", "buttonSEL", "hideSelectedVertices", QIcon(dlg_selected), Qt.ArrowCursor, 0), + ("Hide unselected", "buttonUN", "hideUnSelectedVertices", QIcon(dlg_unselected), Qt.ArrowCursor, 0), + ("Show all nodes", "buttonSW", "showAllVertices", QIcon(dlg_showall), Qt.ArrowCursor, 0))) + + OWGUI.rubber(self.zoomSelectToolbar) + + ib = OWGUI.widgetBox(self.infoTab, "General") + OWGUI.label(ib, self, "Number of vertices: %(nVertices)i") + OWGUI.label(ib, self, "Number of edges: %(nEdges)i") + OWGUI.label(ib, self, "Vertices per edge: %(verticesPerEdge).2f") + OWGUI.label(ib, self, "Edges per vertex: %(edgesPerVertex).2f") + OWGUI.label(ib, self, "Diameter: %(diameter)i") + OWGUI.label(ib, self, "Clustering Coefficient: %(clustering_coefficient).1f%%") + + ib = OWGUI.widgetBox(self.infoTab, orientation="horizontal") + + OWGUI.button(ib, self, "Degree distribution", callback=self.showDegreeDistribution, debuggingEnabled=False) + OWGUI.button(ib, self, "Save network", callback=self.saveNetwork, debuggingEnabled=False) + OWGUI.button(ib, self, "Save image", callback=self.networkCanvas.saveToFile, debuggingEnabled=False) + + #OWGUI.button(self.edgesTab, self, "Clustering", callback=self.clustering) + + ib = OWGUI.widgetBox(self.infoTab, "Prototype") + #OWGUI.button(ib, self, "Collapse", callback=self._collapse) + + #ib = OWGUI.widgetBox(ibProto, "Name components") + OWGUI.lineEdit(ib, self, "organism", "Organism:", orientation='horizontal') + + self.nameComponentAttribute = 0 + self.nameComponentCombo = OWGUI.comboBox(ib, self, "nameComponentAttribute", callback=self.nameComponents, label="Name components:", orientation="horizontal") + self.nameComponentCombo.addItem("Select attribute") + + self.showComponentAttribute = 0 + self.showComponentCombo = OWGUI.comboBox(ib, self, "showComponentAttribute", callback=self.showComponents, label="Labels on components:", orientation="horizontal") + self.showComponentCombo.addItem("Select attribute") + OWGUI.checkBox(ib, self, 'showTextMiningInfo', "Show text mining info") + + #ib = OWGUI.widgetBox(ibProto, "Distance Matrix") + ibs = OWGUI.widgetBox(ib, orientation="horizontal") + self.btnMDS = OWGUI.button(ibs, self, "Fragviz", callback=self.mdsComponents, toggleButton=1) + self.btnESIM = OWGUI.button(ibs, self, "eSim", callback=self.exactSimulation, toggleButton=1) + self.btnMDSv = OWGUI.button(ibs, self, "MDS", callback=self.mdsVertices, toggleButton=1) + ibs = OWGUI.widgetBox(ib, orientation="horizontal") + self.btnRotate = OWGUI.button(ibs, self, "Rotate", callback=self.rotateComponents, toggleButton=1) + self.btnRotateMDS = OWGUI.button(ibs, self, "Rotate (MDS)", callback=self.rotateComponentsMDS, toggleButton=1) + self.btnForce = OWGUI.button(ibs, self, "Draw Force", callback=self.drawForce, toggleButton=1) + self.scalingRatio = 0 + OWGUI.spin(ib, self, "scalingRatio", 0, 9, 1, label="Set scalingRatio: ") + OWGUI.doubleSpin(ib, self, "mdsStressDelta", 0, 10, 0.0000000000000001, label="Min stress change: ") + OWGUI.spin(ib, self, "mdsSteps", 1, 100000, 1, label="MDS steps: ") + OWGUI.spin(ib, self, "mdsRefresh", 1, 100000, 1, label="MDS refresh steps: ") + ibs = OWGUI.widgetBox(ib, orientation="horizontal") + OWGUI.checkBox(ibs, self, 'mdsTorgerson', "Torgerson's approximation") + OWGUI.checkBox(ibs, self, 'mdsAvgLinkage', "Use average linkage") + OWGUI.checkBox(ib, self, 'mdsFromCurrentPos', "MDS from current positions") + self.mdsInfoA=OWGUI.widgetLabel(ib, "Avg. stress:") + self.mdsInfoB=OWGUI.widgetLabel(ib, "Num. steps:") + self.rotateSteps = 100 + + OWGUI.spin(ib, self, "rotateSteps", 1, 10000, 1, label="Rotate max steps: ") + OWGUI.spin(ib, self, "minComponentEdgeWidth", 0, 100, 1, label="Min component edge width: ", callback=self.setComponentEdgeWidth) + OWGUI.spin(ib, self, "maxComponentEdgeWidth", 0, 200, 1, label="Max component edge width: ", callback=self.setComponentEdgeWidth) + + self.attSelectionAttribute = 0 + self.comboAttSelection = OWGUI.comboBox(ib, self, "attSelectionAttribute", label='Send attribute selection list:', orientation='horizontal', callback=self.sendAttSelectionList) + self.comboAttSelection.addItem("Select attribute") + self.autoSendAttributes = 0 + OWGUI.checkBox(ib, self, 'autoSendAttributes', "auto send attributes", callback=self.setAutoSendAttributes) + + self.icons = self.createAttributeIconDict() + self.setMarkMode() + + self.editAttribute = 0 + self.editCombo = OWGUI.comboBox(self.infoTab, self, "editAttribute", label="Edit attribute:", orientation="horizontal") + self.editCombo.addItem("Select attribute") + self.editValue = '' + OWGUI.lineEdit(self.infoTab, self, "editValue", "Value:", orientation='horizontal') + OWGUI.button(self.infoTab, self, "Edit", callback=self.edit) + + self.verticesTab.layout().addStretch(1) + self.edgesTab.layout().addStretch(1) + self.markTab.layout().addStretch(1) + self.infoTab.layout().addStretch(1) + + dlg = self.createColorDialog(self.colorSettings, self.selectedSchemaIndex) + self.networkCanvas.contPalette = dlg.getContinuousPalette("contPalette") + self.networkCanvas.discPalette = dlg.getDiscretePalette("discPalette") + + dlg = self.createColorDialog(self.edgeColorSettings, self.selectedEdgeSchemaIndex) + self.networkCanvas.contEdgePalette = dlg.getContinuousPalette("contPalette") + self.networkCanvas.discEdgePalette = dlg.getDiscretePalette("discPalette") + + self.setOptMethod() + self.setFontSize() + self.tabs.setCurrentIndex(self.tabIndex) + self.setGraph(None) + self.setMinimumWidth(900) + + #self.resize(1000, 600) + #self.controlArea.setEnabled(False) + + def currentTabChanged(self, index): + self.tabIndex = index + + def setComponentEdgeWidth(self): + self.networkCanvas.minComponentEdgeWidth = self.minComponentEdgeWidth + self.networkCanvas.maxComponentEdgeWidth = self.maxComponentEdgeWidth + self.updateCanvas() + + def setAutoSendAttributes(self): + if self.autoSendAttributes: + self.networkCanvas.callbackSelectVertex = self.sendAttSelectionList + else: + self.networkCanvas.callbackSelectVertex = None + + def sendAttSelectionList(self): + if self.optimization: + vars = [x.name for x in self.optimization.getVars()] + if not self.comboAttSelection.currentText() in vars: + return + att = str(self.comboAttSelection.currentText()) + vertices = self.networkCanvas.networkCurve.getSelectedVertices() + + if len(vertices) != 1: + return + + attributes = str(self.optimization.graph.items[vertices[0]][att]).split(', ') + else: + attributes = None + self.send("Features", attributes) + + def edit(self): + if self.optimization is None: + return + + vars = [x.name for x in self.optimization.getVars()] + if not self.editCombo.currentText() in vars: + return + att = str(self.editCombo.currentText()) + vertices = self.networkCanvas.networkCurve.getSelectedVertices() + + if len(vertices) == 0: + return + + if self.optimization.graph.items.domain[att].varType == orange.VarTypes.Continuous: + for v in vertices: + self.optimization.graph.items[v][att] = float(self.editValue) + else: + for v in vertices: + self.optimization.graph.items[v][att] = str(self.editValue) + + self.setItems(self.optimization.graph.items) + + def drawForce(self): + if self.btnForce.isChecked() and self.optimization is not None: + self.networkCanvas.forceVectors = self.optimization._computeForces() + else: + self.networkCanvas.forceVectors = None + + self.updateCanvas() + + def rotateProgress(self, curr, max): + self.progressBarSet(int(curr * 100 / max)) + qApp.processEvents() + + def rotateComponentsMDS(self): + print "rotate" + if self.vertexDistance is None: + self.information('Set distance matrix to input signal') + self.btnRotateMDS.setChecked(False) + return + + if self.optimization is None: + self.information('No network found') + self.btnRotateMDS.setChecked(False) + return + if self.vertexDistance.dim != self.optimization.graph.nVertices: + self.error('Distance matrix dimensionality must equal number of vertices') + self.btnRotateMDS.setChecked(False) + return + + if not self.btnRotateMDS.isChecked(): + self.optimization.stopMDS = 1 + #self.btnMDS.setChecked(False) + #self.btnMDS.setText("MDS on graph components") + return + + self.btnRotateMDS.setText("Stop") + qApp.processEvents() + + self.optimization.vertexDistance = self.vertexDistance + self.progressBarInit() + + self.optimization.mdsComponents(self.mdsSteps, self.mdsRefresh, self.mdsProgress, self.updateCanvas, self.mdsTorgerson, self.mdsStressDelta, rotationOnly=True, mdsFromCurrentPos=self.mdsFromCurrentPos) + + self.btnRotateMDS.setChecked(False) + self.btnRotateMDS.setText("Rotate graph components (MDS)") + self.progressBarFinished() + + def rotateComponents(self): + if self.vertexDistance is None: + self.information('Set distance matrix to input signal') + self.btnRotate.setChecked(False) + return + + if self.optimization is None: + self.information('No network found') + self.btnRotate.setChecked(False) + return + + if self.vertexDistance.dim != self.optimization.graph.nVertices: + self.error('Distance matrix dimensionality must equal number of vertices') + self.btnRotate.setChecked(False) + return + + if not self.btnRotate.isChecked(): + self.optimization.stopRotate = 1 + return + + self.btnRotate.setText("Stop") + qApp.processEvents() + + self.optimization.vertexDistance = self.vertexDistance + self.progressBarInit() + self.optimization.rotateComponents(self.rotateSteps, 0.0001, self.rotateProgress, self.updateCanvas) + self.btnRotate.setChecked(False) + self.btnRotate.setText("Rotate graph components") + self.progressBarFinished() + + def mdsProgress(self, avgStress, stepCount): + self.drawForce() + + self.mdsInfoA.setText("Avg. Stress: %.20f" % avgStress) + self.mdsInfoB.setText("Num. steps: %i" % stepCount) + self.progressBarSet(int(stepCount * 100 / self.mdsSteps)) + qApp.processEvents() + + def mdsComponents(self, mdsType=orngNetwork.MdsType.componentMDS): + if mdsType == orngNetwork.MdsType.componentMDS: + btn = self.btnMDS + elif mdsType == orngNetwork.MdsType.exactSimulation: + btn = self.btnESIM + elif mdsType == orngNetwork.MdsType.MDS: + btn = self.btnMDSv + + btnCaption = btn.text() + + if self.vertexDistance is None: + self.information('Set distance matrix to input signal') + btn.setChecked(False) + return + + if self.optimization is None: + self.information('No network found') + btn.setChecked(False) + return + + if self.vertexDistance.dim != self.optimization.graph.nVertices: + self.error('Distance matrix dimensionality must equal number of vertices') + btn.setChecked(False) + return + + if not btn.isChecked(): + self.optimization.stopMDS = 1 + btn.setChecked(False) + btn.setText(btnCaption) + return + + btn.setText("Stop") + qApp.processEvents() + + self.optimization.vertexDistance = self.vertexDistance + self.progressBarInit() + + if self.mdsAvgLinkage: + self.optimization.mdsComponentsAvgLinkage(self.mdsSteps, self.mdsRefresh, self.mdsProgress, self.updateCanvas, self.mdsTorgerson, self.mdsStressDelta, scalingRatio = self.scalingRatio, mdsFromCurrentPos=self.mdsFromCurrentPos) + else: + self.optimization.mdsComponents(self.mdsSteps, self.mdsRefresh, self.mdsProgress, self.updateCanvas, self.mdsTorgerson, self.mdsStressDelta, mdsType=mdsType, scalingRatio=self.scalingRatio, mdsFromCurrentPos=self.mdsFromCurrentPos) + + btn.setChecked(False) + btn.setText(btnCaption) + self.progressBarFinished() + + def exactSimulation(self): + self.mdsComponents(orngNetwork.MdsType.exactSimulation) + + def mdsVertices(self): + self.mdsComponents(orngNetwork.MdsType.MDS) + + def setVertexDistance(self, matrix): + self.error('') + self.information('') + + if matrix is None or self.optimization is None or self.optimization.graph is None: + self.vertexDistance = None + if self.optimization: self.optimization.vertexDistance = None + if self.networkCanvas: self.networkCanvas.vertexDistance = None + return + + if matrix.dim != self.optimization.graph.nVertices: + self.error('Distance matrix dimensionality must equal number of vertices') + self.vertexDistance = None + if self.optimization: self.optimization.vertexDistance = None + if self.networkCanvas: self.networkCanvas.vertexDistance = None + return + + self.vertexDistance = matrix + if self.optimization: self.optimization.vertexDistance = matrix + if self.networkCanvas: self.networkCanvas.vertexDistance = matrix + + self.updateCanvas() + + def setSendMarkedNodes(self): + if self.checkSendMarkedNodes: + self.networkCanvas.sendMarkedNodes = self.sendMarkedNodes + self.sendMarkedNodes(self.networkCanvas.getMarkedVertices()) + else: + self.send("Marked Data", None) + self.networkCanvas.sendMarkedNodes = None + + def sendMarkedNodes(self, markedNodes): + if len(markedNodes) == 0: + self.send("Marked Data", None) + return + + if self.optimization != None and self.optimization.graph != None and self.optimization.graph.items != None: + items = self.optimization.graph.items.getitems(markedNodes) + self.send("Marked Data", items) + return + + self.send("Marked Data", None) + + def _collapse(self): + #print "collapse" + self.optimization.collapse() + self.networkCanvas.addVisualizer(self.optimization) + #if not nodes is None: + # self.networkCanvas.updateData() + # self.networkCanvas.addSelection(nodes, False) + self.updateCanvas() + + def clustering(self): + #print "clustering" + self.optimization.graph.getClusters() + + def insideviewneighbours(self): + if self.networkCanvas.insideview == 1: + self.networkCanvas.insideviewNeighbours = self.insideViewNeighbours + self.optButton.setChecked(True) + self.fr(False) + + def insideview(self): + print self.networkCanvas.getSelectedVertices() + if len(self.networkCanvas.getSelectedVertices()) == 1: + if self.networkCanvas.insideview == 1: + print "insideview: 1" + self.networkCanvas.insideview = 0 + self.networkCanvas.showAllVertices() + self.updateCanvas() + else: + print "insideview: 0" + self.networkCanvas.insideview = 1 + self.networkCanvas.insideviewNeighbors = self.insideViewNeighbours + self.optButton.setChecked(True) + self.fr(False) + + else: + print "One node must be selected!" + + def showComponents(self): + if self.optimization is None or self.optimization.graph is None or self.optimization.graph.items is None: + return + + vars = [x.name for x in self.optimization.getVars()] + + if not self.showComponentCombo.currentText() in vars: + self.networkCanvas.showComponentAttribute = None + self.lastNameComponentAttribute = '' + else: + self.networkCanvas.showComponentAttribute = self.showComponentCombo.currentText() + + self.networkCanvas.drawPlotItems() + + def nameComponents(self): + """Names connected components of genes according to GO terms.""" + self.progressBarFinished() + self.lastNameComponentAttribute = None + + if self.optimization is None or self.optimization.graph is None or self.optimization.graph.items is None: + return + + vars = [x.name for x in self.optimization.getVars()] + if not self.nameComponentCombo.currentText() in vars: + return + + self.progressBarInit() + components = [c for c in self.optimization.graph.getConnectedComponents() if len(c) > 1] + if 'component name' in self.optimization.graph.items.domain: + keyword_table = self.optimization.graph.items + else: + keyword_table = orange.ExampleTable(orange.Domain(orange.StringVariable('component name')), [[''] for i in range(len(self.optimization.graph.items))]) + + import obiGO + ontology = obiGO.Ontology.Load(progressCallback=self.progressBarSet) + annotations = obiGO.Annotations.Load(self.organism, ontology=ontology, progressCallback=self.progressBarSet) + + allGenes = set([e[str(self.nameComponentCombo.currentText())].value for e in self.optimization.graph.items]) + foundGenesets = False + if len(annotations.geneNames & allGenes) < 1: + allGenes = set(reduce(operator.add, [e[str(self.nameComponentCombo.currentText())].value.split(', ') for e in self.optimization.graph.items])) + if len(annotations.geneNames & allGenes) < 1: + self.warning('no genes found') + return + else: + foundGenesets = True + + def rank(a, j, reverse=False): + if len(a) <= 0: return + + if reverse: + a.sort(lambda x, y: 1 if x[j] > y[j] else -1 if x[j] < y[j] else 0) + top_value = a[0][j] + top_rank = len(a) + max_rank = float(len(a)) + int_ndx = 0 + for k in range(len(a)): + if top_value < a[k][j]: + top_value = a[k][j] + if k - int_ndx > 1: + avg_rank = (a[int_ndx][j] + a[k-1][j]) / 2 + for l in range(int_ndx, k): + a[l][j] = avg_rank + + int_ndx = k + + a[k][j] = top_rank / max_rank + top_rank -= 1 + + k += 1 + if k - int_ndx > 1: + avg_rank = (a[int_ndx][j] + a[k-1][j]) / 2 + for l in range(int_ndx, k): + a[l][j] = avg_rank + + else: + a.sort(lambda x, y: 1 if x[j] < y[j] else -1 if x[j] > y[j] else 0) + top_value = a[0][j] + top_rank = len(a) + max_rank = float(len(a)) + int_ndx = 0 + for k in range(len(a)): + if top_value > a[k][j]: + top_value = a[k][j] + if k - int_ndx > 1: + avg_rank = (a[int_ndx][j] + a[k-1][j]) / 2 + for l in range(int_ndx, k): + a[l][j] = avg_rank + + int_ndx = k + + a[k][j] = top_rank / max_rank + top_rank -= 1 + + k += 1 + if k - int_ndx > 1: + avg_rank = (a[int_ndx][j] + a[k-1][j]) / 2 + for l in range(int_ndx, k): + a[l][j] = avg_rank + + for i in range(len(components)): + component = components[i] + if len(component) <= 1: + continue + + if foundGenesets: + genes = reduce(operator.add, [self.optimization.graph.items[v][str(self.nameComponentCombo.currentText())].value.split(', ') for v in component]) + else: + genes = [self.optimization.graph.items[v][str(self.nameComponentCombo.currentText())].value for v in component] + + res1 = annotations.GetEnrichedTerms(genes, aspect="P") + res2 = annotations.GetEnrichedTerms(genes, aspect="F") + res = res1.items() + res2.items() + #namingScore = [[(1-p_value) * (float(len(g)) / len(genes)) / (float(ref) / len(annotations.geneNames)), ontology.terms[GOId].name, len(g), ref, p_value] for GOId, (g, p_value, ref) in res.items() if p_value < 0.1] + #namingScore = [[(1-p_value) * len(g) / ref, ontology.terms[GOId].name, len(g), ref, p_value] for GOId, (g, p_value, ref) in res.items() if p_value < 0.1] + + namingScore = [[len(g), ref, p_value, ontology[GOId].name, len(g), ref, p_value] for GOId, (g, p_value, ref) in res if p_value < 0.1] + if len(namingScore) == 0: + continue + + annotated_genes = max([a[0] for a in namingScore]) + + rank(namingScore, 1, reverse=True) + rank(namingScore, 2, reverse=True) + rank(namingScore, 0) + + namingScore = [[10 * rank_genes + 0.5 * rank_ref + rank_p_value, name, g, ref, p_value] for rank_genes, rank_ref, rank_p_value, name, g, ref, p_value in namingScore] + namingScore.sort(reverse=True) + + if len(namingScore) < 1: + print "warning. no annotations found for group of genes: " + ", ".join(genes) + continue + elif len(namingScore[0]) < 2: + print "warning. error computing score for group of genes: " + ", ".join(genes) + continue + + for v in component: + name = str(namingScore[0][1]) + attrs = "%d/%d, %d, %lf" % (namingScore[0][2], annotated_genes, namingScore[0][3], namingScore[0][4]) + info = '' + if self.showTextMiningInfo: + info = "\n" + attrs + "\n" + str(namingScore[0][0]) + keyword_table[v]['component name'] = name + info + + self.progressBarSet(i*100.0/len(components)) + + self.lastNameComponentAttribute = self.nameComponentCombo.currentText() + self.setItems(orange.ExampleTable([self.optimization.graph.items, keyword_table])) + self.progressBarFinished() + + def nameComponents_old(self): + if self.optimization is None or self.optimization.graph is None or self.optimization.graph.items is None: + return + + vars = [x.name for x in self.optimization.getVars()] + + if not self.nameComponentCombo.currentText() in vars: + return + + components = self.optimization.graph.getConnectedComponents() + keyword_table = orange.ExampleTable(orange.Domain(orange.StringVariable('component name')), [[''] for i in range(len(self.optimization.graph.items))]) + + excludeWord = ["AND", "OF", "KEGG", "ST", "IN", "SIG"] + excludePart = ["HSA"] + keywords = set() + sameKeywords = set() + + for component in components: + words = [] + all_values = [] + for vertex in component: + values = [] + value = str(self.optimization.graph.items[vertex][str(self.nameComponentCombo.currentText())]) + + value = value.replace(" ", ",") + value_top = value.split(",") + + for value in value_top: + if len(value) > 0: + tmp = value.split("_") + tmp = [value.strip() for value in tmp if len(value) > 0] + all_values.append(tmp) + values.extend(tmp) + + values = [value.strip() for value in values if len(value) > 0] + words.extend(values) + + + #value = str(self.optimization.graph.items[vertex][str(self.nameComponentCombo.currentText())]) + #value = value.replace(" ", "_") + #value = value.replace(",", "_") + #values = value.split("_") + #values = [value.strip() for value in values if len(value) > 0] + #print "values:", values + #all_values.append(values) + + #words.extend(values) + #print "all_values:", all_values + toExclude = [] + + words = [word for word in words if not word.upper() in excludeWord] + toExclude = [word for word in words for part in excludePart if word.find(part) != -1] + + for word in toExclude: + try: + while True: + words.remove(word) + except: + pass + + counted_words = {} + for word in words: + if word in counted_words: + count = counted_words[word] + counted_words[word] = count + 1 + else: + counted_words[word] = 1 + + words = sorted(counted_words.items(), key=itemgetter(1), reverse=True) + keyword = "" + keyword_words = [] + max_count = 0 + i = 0 + + while i < len(words) and words[i][1] >= max_count: + max_count = words[i][1] + keyword += words[i][0] + " " + keyword_words.append(words[i][0]) + i += 1 + + if len(keyword_words) > 1: + new_all_values = [] + for values in all_values: + new_values = [value for value in values if value in keyword_words] + new_all_values.append(new_values) + + #print new_all_values + word_position = [] + + for word in keyword_words: + sum = 0 + for v in new_all_values: + if word in v: + sum += v.index(word) + + word_position.append((word, sum)) + + words = sorted(word_position, key=itemgetter(1)) + #print "words:", words + #print all_values + #print new_all_values + + keyword = "" + for word in words: + keyword += word[0] + " " + + keyword = keyword.strip() + + for vertex in component: + keyword_table[vertex]['component name'] = keyword + + if keyword in keywords: + sameKeywords.add(keyword) + else: + keywords.add(keyword) + #print "sameKeywords:", sameKeywords + sameComponents = [component for component in components if str(keyword_table[component[0]]['component name']) in sameKeywords] + #print "same components:", sameComponents + + for component in sameComponents: + words = [] + all_values = [] + for vertex in component: + values = [] + value = str(self.optimization.graph.items[vertex][str(self.nameComponentCombo.currentText())]) + + value = value.replace(" ", ",") + value_top = value.split(",") + + for value in value_top: + if len(value) > 0: + tmp = value.split("_") + tmp = [value.strip() for value in tmp if len(value) > 0] + all_values.append(tmp) + values.extend(tmp) + + values = [value.strip() for value in values if len(value) > 0] + words.extend(values) + + toExclude = [] + + words = [word for word in words if not word.upper() in excludeWord] + toExclude = [word for word in words for part in excludePart if word.find(part) != -1] + + for word in toExclude: + try: + while True: + words.remove(word) + except: + pass + + counted_words = {} + for word in words: + if word in counted_words: + count = counted_words[word] + counted_words[word] = count + 1 + else: + counted_words[word] = 1 + + words = sorted(counted_words.items(), key=itemgetter(1), reverse=True) + keyword = "" + counts = [int(word[1]) for word in words] + max_count = max(counts) + + try: + while True and len(counts) > 1: + counts.remove(max_count) + except: + pass + max_count = max(counts) + i = 0 + keyword_words = [] + while i < len(words) and words[i][1] >= max_count: + keyword += words[i][0] + " " + keyword_words.append(words[i][0]) + i += 1 + + if len(keyword_words) > 1: + new_all_values = [] + for values in all_values: + new_values = [value for value in values if value in keyword_words] + new_all_values.append(new_values) + + #print new_all_values + word_position = [] + + for word in keyword_words: + sum = 0 + for v in new_all_values: + if word in v: + sum += v.index(word) + + word_position.append((word, sum)) + + words = sorted(word_position, key=itemgetter(1)) + #print "words:", words + #print all_values + #print new_all_values + + keyword = "" + for word in words: + keyword += word[0] + " " + + keyword = keyword.strip() + for vertex in component: + keyword_table[vertex]['component name'] = keyword + + self.lastNameComponentAttribute = self.nameComponentCombo.currentText() + #print "self.lastNameComponentAttribute:", self.lastNameComponentAttribute + items = orange.ExampleTable([self.optimization.graph.items, keyword_table]) + self.setItems(items) + + #for item in items: + # print item + + def showIndexLabels(self): + self.networkCanvas.showIndexes = self.showIndexes + self.networkCanvas.updateData() + self.networkCanvas.replot() + + def showWeightLabels(self): + self.networkCanvas.showWeights = self.showWeights + self.networkCanvas.updateData() + self.networkCanvas.replot() + + def showDistancesClick(self): + if self.visualize and self.visualize.vertexDistance is None: + self.warning("Vertex distance signal is not set. Distances are not known.") + self.networkCanvas.showDistances = self.showDistances + + def showEdgeLabelsClick(self): + self.networkCanvas.showEdgeLabels = self.showEdgeLabels + self.networkCanvas.updateData() + self.networkCanvas.replot() + + def labelsOnMarked(self): + self.networkCanvas.labelsOnMarkedOnly = self.labelsOnMarkedOnly + self.networkCanvas.updateData() + self.networkCanvas.replot() + + def setSearchStringTimer(self): + self.hubs = 1 + self.searchStringTimer.stop() + self.searchStringTimer.start(1000) + + def setMarkMode(self, i = None): + self.searchStringTimer.stop() + if not i is None: + self.hubs = i + + #print self.hubs + self.networkCanvas.tooltipNeighbours = self.hubs == 2 and self.markDistance or 0 + self.networkCanvas.markWithRed = False + + if not self.optimization or not self.optimization.graph: + return + + hubs = self.hubs + vgraph = self.optimization.graph + + if hubs == 0: + self.networkCanvas.setMarkedVertices([]) + self.networkCanvas.replot() + return + + elif hubs == 1: + #print "mark on given label" + txt = self.markSearchString + labelText = self.networkCanvas.labelText + self.networkCanvas.markWithRed = self.networkCanvas.nVertices > 200 + + toMark = [i for i, values in enumerate(vgraph.items) if txt.lower() in " ".join([str(values[ndx]).decode("ascii", "ignore").lower() for ndx in range(len(vgraph.items.domain)) + vgraph.items.domain.getmetas().keys()])] + self.networkCanvas.setMarkedVertices(toMark) + self.networkCanvas.replot() + return + + elif hubs == 2: + #print "mark on focus" + self.networkCanvas.unMark() + self.networkCanvas.tooltipNeighbours = self.markDistance + return + + elif hubs == 3: + #print "mark selected" + self.networkCanvas.unMark() + self.networkCanvas.selectionNeighbours = self.markDistance + self.networkCanvas.markSelectionNeighbours() + return + + self.networkCanvas.tooltipNeighbours = self.networkCanvas.selectionNeighbours = 0 + powers = vgraph.getDegrees() + + if hubs == 4: # at least N connections + #print "mark at least N connections" + N = self.markNConnections + self.networkCanvas.setMarkedVertices([i for i, power in enumerate(powers) if power >= N]) + self.networkCanvas.replot() + elif hubs == 5: + #print "mark at most N connections" + N = self.markNConnections + self.networkCanvas.setMarkedVertices([i for i, power in enumerate(powers) if power <= N]) + self.networkCanvas.replot() + elif hubs == 6: + #print "mark more than any" + self.networkCanvas.setMarkedVertices([i for i, power in enumerate(powers) if power > max([0]+[powers[nn] for nn in vgraph.getNeighbours(i)])]) + self.networkCanvas.replot() + elif hubs == 7: + #print "mark more than avg" + self.networkCanvas.setMarkedVertices([i for i, power in enumerate(powers) if power > mean([0]+[powers[nn] for nn in vgraph.getNeighbours(i)])]) + self.networkCanvas.replot() + elif hubs == 8: + #print "mark most" + sortedIdx = range(len(powers)) + sortedIdx.sort(lambda x,y: -cmp(powers[x], powers[y])) + cutP = self.markNumber - 1 + cutPower = powers[sortedIdx[cutP]] + while cutP < len(powers) and powers[sortedIdx[cutP]] == cutPower: + cutP += 1 + self.networkCanvas.setMarkedVertices(sortedIdx[:cutP]) + self.networkCanvas.replot() + elif hubs == 9: + self.setMarkInput() + + def testRefresh(self): + start = time() + self.networkCanvas.replot() + stop = time() + print "replot in " + str(stop - start) + + def saveNetwork(self): + if self.networkCanvas is None or self.networkCanvas.visualizer is None: + return + + filename = QFileDialog.getSaveFileName(self, 'Save Network File', '', 'PAJEK network (*.net)\nGML network (*.gml)') + filename = unicode(filename) + if filename: + fn = "" + head, tail = os.path.splitext(filename) + if not tail: + fn = head + ".net" + else: + fn = filename + + self.optimization.graph.save(fn) + + def sendData(self): + graph = self.networkCanvas.getSelectedGraph() + vertices = self.networkCanvas.getSelectedVertices() + + if graph != None: + if graph.items != None: + self.send("Selected Data", graph.items) + else: + self.send("Selected Data", self.networkCanvas.getSelectedExamples()) + + #print "sendData:", self.visualize.graph.items.domain + self.send("Other Data", self.networkCanvas.getUnselectedExamples()) + self.send("Selected Network", graph) + else: + items = self.networkCanvas.getSelectedExamples() + self.send("Selected Data", items) + + items = self.networkCanvas.getUnselectedExamples() + self.send("Other Data", items) + + matrix = None + if self.vertexDistance != None: + matrix = self.vertexDistance.getitems(vertices) + + self.send("Selected Distance Matrix", matrix) + + def setCombos(self): + vars = self.optimization.getVars() + edgeVars = self.optimization.getEdgeVars() + lastLabelColumns = self.lastLabelColumns + lastTooltipColumns = self.lastTooltipColumns + + self.clearCombos() + + self.attributes = [(var.name, var.varType) for var in vars] + self.edgeAttributes = [(var.name, var.varType) for var in edgeVars] + + for var in vars: + if var.varType in [orange.VarTypes.Discrete, orange.VarTypes.Continuous]: + self.colorCombo.addItem(self.icons[var.varType], unicode(var.name)) + + if var.varType in [orange.VarTypes.String] and hasattr(self.optimization.graph, 'items') and self.optimization.graph.items != None and len(self.optimization.graph.items) > 0: + + value = self.optimization.graph.items[0][var].value + + # can value be a list? + try: + if type(eval(value)) == type([]): + self.vertexSizeCombo.addItem(self.icons[var.varType], unicode(var.name)) + continue + except: + pass + + if len(value.split(',')) > 1: + self.vertexSizeCombo.addItem(self.icons[var.varType], "num of " + unicode(var.name)) + + elif var.varType in [orange.VarTypes.Continuous]: + self.vertexSizeCombo.addItem(self.icons[var.varType], unicode(var.name)) + + self.nameComponentCombo.addItem(self.icons[var.varType], unicode(var.name)) + self.showComponentCombo.addItem(self.icons[var.varType], unicode(var.name)) + self.editCombo.addItem(self.icons[var.varType], unicode(var.name)) + self.comboAttSelection.addItem(self.icons[var.varType], unicode(var.name)) + + for var in edgeVars: + if var.varType in [orange.VarTypes.Discrete, orange.VarTypes.Continuous]: + self.edgeColorCombo.addItem(self.icons[var.varType], unicode(var.name)) + + for i in range(self.vertexSizeCombo.count()): + if self.lastVertexSizeColumn == self.vertexSizeCombo.itemText(i): + self.vertexSize = i + break + + for i in range(self.colorCombo.count()): + if self.lastColorColumn == self.colorCombo.itemText(i): + self.color = i + break + + for i in range(self.attListBox.count()): + if str(self.attListBox.item(i).text()) in lastLabelColumns: + self.attListBox.item(i).setSelected(1) + + for i in range(self.tooltipListBox.count()): + if str(self.tooltipListBox.item(i).text()) in lastTooltipColumns: + self.tooltipListBox.item(i).setSelected(1) + + self.lastLabelColumns = lastLabelColumns + self.lastTooltipColumns = lastTooltipColumns + + def clearCombos(self): + self.attributes = [] + self.edgeAttributes = [] + + self.colorCombo.clear() + self.vertexSizeCombo.clear() + self.nameComponentCombo.clear() + self.showComponentCombo.clear() + self.edgeColorCombo.clear() + self.editCombo.clear() + self.comboAttSelection.clear() + + self.colorCombo.addItem("(same color)") + self.edgeColorCombo.addItem("(same color)") + self.vertexSizeCombo.addItem("(same size)") + self.nameComponentCombo.addItem("Select attribute") + self.showComponentCombo.addItem("Select attribute") + self.editCombo.addItem("Select attribute") + self.comboAttSelection.addItem("Select attribute") + + def setGraph(self, graph): + if graph is None: + self.optimization = None + self.networkCanvas.addVisualizer(self.optimization) + self.clearCombos() + return + + #print "OWNetwork/setGraph: new visualizer..." + self.optimization = orngNetwork.NetworkOptimization(graph) + self.networkCanvas.addVisualizer(self.optimization) + self.networkCanvas.renderAntialiased = self.renderAntialiased + self.networkCanvas.showEdgeLabels = self.showEdgeLabels + self.networkCanvas.minVertexSize = self.minVertexSize + self.networkCanvas.maxVertexSize = self.maxVertexSize + self.networkCanvas.maxEdgeSize = self.maxLinkSize + self.lastVertexSizeColumn = self.vertexSizeCombo.currentText() + self.lastColorColumn = self.colorCombo.currentText() + self.networkCanvas.minComponentEdgeWidth = self.minComponentEdgeWidth + self.networkCanvas.maxComponentEdgeWidth = self.maxComponentEdgeWidth + + + + #for i in range(graph.nVertices): + # print "x:", graph.coors[0][i], " y:", graph.coors[1][i] + + self.nVertices = graph.nVertices + self.nShown = graph.nVertices + + if graph.directed: + self.nEdges = len(graph.getEdges()) + else: + self.nEdges = len(graph.getEdges()) / 2 + + if self.nEdges > 0: + self.verticesPerEdge = float(self.nVertices) / float(self.nEdges) + else: + self.verticesPerEdge = 0 + + if self.nVertices > 0: + self.edgesPerVertex = float(self.nEdges) / float(self.nVertices) + else: + self.edgesPerVertex = 0 + + self.diameter = graph.getDiameter() + self.clustering_coefficient = graph.getClusteringCoefficient() * 100 + + self.setCombos() + + lastNameComponentAttributeFound = False + for i in range(self.nameComponentCombo.count()): + if self.lastNameComponentAttribute == self.nameComponentCombo.itemText(i): + lastNameComponentAttributeFound = True + self.nameComponentAttribute = i + self.nameComponents() + self.showComponentAttribute = self.showComponentCombo.count() - 1 + self.showComponents() + break + + if not lastNameComponentAttributeFound: + self.lastNameComponentAttribute = '' + + self.showComponentAttribute = None + + k = 1.13850193174e-008 + nodes = self.optimization.nVertices() + t = k * nodes * nodes + self.frSteps = int(5.0 / t) + if self.frSteps < 1: self.frSteps = 1; + if self.frSteps > 1500: self.frSteps = 1500; + + self.networkCanvas.labelsOnMarkedOnly = self.labelsOnMarkedOnly + self.networkCanvas.showWeights = self.showWeights + self.networkCanvas.showIndexes = self.showIndexes + # if graph is large, set random layout, min vertex size, min edge size + if self.frSteps < 10: + self.renderAntialiased = 0 + self.minVertexSize = 5 + self.maxVertexSize = 5 + self.maxLinkSize = 1 + self.optMethod = 0 + self.setOptMethod() + + if self.vertexSize > 0: + self.networkCanvas.setVerticesSize(self.vertexSizeCombo.currentText(), self.invertSize) + else: + self.networkCanvas.setVerticesSize() + + self.setVertexColor() + self.setEdgeColor() + + self.networkCanvas.setEdgesSize() + self.clickedAttLstBox() + self.clickedTooltipLstBox() + self.clickedEdgeLabelListBox() + + self.optButton.setChecked(1) + self.optLayout() + self.information(0) + #self.controlArea.setEnabled(True) + self.updateCanvas() + + def setItems(self, items=None): + self.error('') + + if self.optimization is None or self.optimization.graph is None or items is None: + return + + if len(items) != self.optimization.graph.nVertices: + self.error('ExampleTable items must have one example for each vertex.') + return + + self.optimization.graph.setattr("items", items) + + self.setVertexSize() + self.showIndexLabels() + self.showWeightLabels() + self.showEdgeLabelsClick() + + self.setCombos() + self.networkCanvas.updateData() + + def setMarkInput(self): + var = str(self.markInputCombo.currentText()) + #print 'combo:',self.markInputCombo.currentText() + if self.markInputItems != None and len(self.markInputItems) > 0: + values = [str(x[var]).strip().upper() for x in self.markInputItems] + toMark = [i for (i,x) in enumerate(self.optimization.graph.items) if str(x[var]).strip().upper() in values] + #print "mark:", toMark + self.networkCanvas.setMarkedVertices(list(toMark)) + self.networkCanvas.replot() + + else: + self.networkCanvas.setMarkedVertices([]) + self.networkCanvas.replot() + + def markItems(self, items): + self.markInputCombo.clear() + self.markInputRadioButton.setEnabled(False) + self.markInputItems = items + + if self.optimization is None or self.optimization.graph is None or self.optimization.graph.items is None or items is None: + return + + if len(items) > 0: + lstOrgDomain = [x.name for x in self.optimization.graph.items.domain] + [self.optimization.graph.items.domain[x].name for x in self.optimization.graph.items.domain.getmetas()] + lstNewDomain = [x.name for x in items.domain] + [items.domain[x].name for x in items.domain.getmetas()] + commonVars = set(lstNewDomain) & set(lstOrgDomain) + + if len(commonVars) > 0: + for var in commonVars: + orgVar = self.optimization.graph.items.domain[var] + mrkVar = items.domain[var] + + if orgVar.varType == mrkVar.varType and orgVar.varType == orange.VarTypes.String: + self.markInputCombo.addItem(self.icons[orgVar.varType], unicode(orgVar.name)) + self.markInputRadioButton.setEnabled(True) + + self.setMarkMode(9) + + def setExampleSubset(self, subset): + if self.networkCanvas is None: + return + + self.warning('') + hiddenNodes = [] + + if subset != None: + try: + expected = 1 + for row in subset: + index = int(row['index'].value) + if expected != index: + hiddenNodes += range(expected-1, index-1) + expected = index + 1 + else: + expected += 1 + + hiddenNodes += range(expected-1, self.networkCanvas.nVertices) + + self.networkCanvas.setHiddenNodes(hiddenNodes) + except: + self.warning('"index" attribute does not exist in "items" table.') + + def updateCanvas(self): + # if network exists + if self.optimization != None: + self.networkCanvas.updateCanvas() + + def showDegreeDistribution(self): + if self.optimization is None: + return + + from matplotlib import rcParams + import pylab as p + + x = self.optimization.graph.getDegrees() + nbins = len(set(x)) + if nbins > 500: + bbins = 500 + #print len(x) + print x + # the histogram of the data + n, bins, patches = p.hist(x, nbins) + + p.xlabel('Degree') + p.ylabel('No. of nodes') + p.title(r'Degree distribution') + + p.show() + + def setColors(self): + dlg = self.createColorDialog(self.colorSettings, self.selectedSchemaIndex) + if dlg.exec_(): + self.colorSettings = dlg.getColorSchemas() + self.selectedSchemaIndex = dlg.selectedSchemaIndex + self.networkCanvas.contPalette = dlg.getContinuousPalette("contPalette") + self.networkCanvas.discPalette = dlg.getDiscretePalette("discPalette") + + self.setVertexColor() + + def setEdgeColorPalette(self): + dlg = self.createColorDialog(self.edgeColorSettings, self.selectedEdgeSchemaIndex) + if dlg.exec_(): + self.edgeColorSettings = dlg.getColorSchemas() + self.selectedEdgeSchemaIndex = dlg.selectedSchemaIndex + self.networkCanvas.contEdgePalette = dlg.getContinuousPalette("contPalette") + self.networkCanvas.discEdgePalette = dlg.getDiscretePalette("discPalette") + + self.setEdgeColor() + + def createColorDialog(self, colorSettings, selectedSchemaIndex): + c = OWColorPalette.ColorPaletteDlg(self, "Color Palette") + c.createDiscretePalette("discPalette", "Discrete Palette") + c.createContinuousPalette("contPalette", "Continuous Palette") + c.setColorSchemas(colorSettings, selectedSchemaIndex) + return c + """ + Layout Optimization + """ + def optLayout(self): + if self.optimization is None: #grafa se ni + self.optButton.setChecked(False) + return + + if not self.optButton.isChecked(): + self.optButton.setChecked(False) + return + + qApp.processEvents() + + if self.optMethod == 1: + self.random() + elif self.optMethod == 2: + self.fr(False) + elif self.optMethod == 3: + self.fr(True) + elif self.optMethod == 4: + self.frRadial() + elif self.optMethod == 5: + self.circularCrossingReduction() + elif self.optMethod == 6: + self.circularOriginal() + elif self.optMethod == 7: + self.circularRandom() + elif self.optMethod == 8: + self.pivotMDS() + + self.optButton.setChecked(False) + qApp.processEvents() + + def setOptMethod(self, method=None): + self.stepsSpin.label.setText('Iterations: ') + + if method != None: + self.optMethod = method + + if str(self.optMethod) == '0': + self.optButton.setEnabled(False) + else: + self.optButton.setEnabled(True) + + if str(self.optMethod) in ['2', '3', '4']: + self.stepsSpin.setEnabled(True) + elif str(self.optMethod) == '8': + self.stepsSpin.label.setText('Pivots: ') + self.stepsSpin.setEnabled(True) + else: + self.stepsSpin.setEnabled(False) + self.optButton.setChecked(True) + self.optLayout() + + def random(self): + #print "OWNetwork/random.." + if self.optimization is None: #grafa se ni + return + + self.optimization.random() + #print self.optimization.coors + #print "OWNetwork/random: updating canvas..." + self.updateCanvas(); + + def fr(self, weighted): + if self.optimization is None: #grafa se ni + return + + if not self.optButton.isChecked(): + #print "not" + self.stopOptimization = 1 + self.optButton.setChecked(False) + self.optButton.setText("Optimize layout") + return + + self.optButton.setText("Stop") + qApp.processEvents() + self.stopOptimization = 0 + tolerance = 5 + initTemp = 1000 + breakpoints = 6 + k = int(self.frSteps / breakpoints) + o = self.frSteps % breakpoints + iteration = 0 + coolFactor = math.exp(math.log(10.0/10000.0) / self.frSteps) + + if k > 0: + while iteration < breakpoints: + #print "iteration, initTemp: " + str(initTemp) + if self.stopOptimization: + return + initTemp = self.optimization.fruchtermanReingold(k, initTemp, coolFactor, self.networkCanvas.hiddenNodes, weighted) + iteration += 1 + qApp.processEvents() + self.updateCanvas() + + #print "ostanek: " + str(o) + ", initTemp: " + str(initTemp) + if self.stopOptimization: + return + initTemp = self.optimization.fruchtermanReingold(o, initTemp, coolFactor, self.networkCanvas.hiddenNodes, weighted) + qApp.processEvents() + self.updateCanvas() + else: + while iteration < o: + #print "iteration ostanek, initTemp: " + str(initTemp) + if self.stopOptimization: + return + initTemp = self.optimization.fruchtermanReingold(1, initTemp, coolFactor, self.networkCanvas.hiddenNodes, weighted) + iteration += 1 + qApp.processEvents() + self.updateCanvas() + + self.optButton.setChecked(False) + self.optButton.setText("Optimize layout") + + def frSpecial(self): + if self.optimization is None: #grafa se ni + return + + steps = 100 + initTemp = 1000 + coolFactor = math.exp(math.log(10.0/10000.0) / steps) + oldXY = [] + for rec in self.optimization.network.coors: + oldXY.append([rec[0], rec[1]]) + #print oldXY + initTemp = self.optimization.fruchtermanReingold(steps, initTemp, coolFactor, self.networkCanvas.hiddenNodes) + #print oldXY + self.networkCanvas.updateDataSpecial(oldXY) + self.networkCanvas.replot() + + def frRadial(self): + if self.optimization is None: #grafa se ni + return + + #print "F-R Radial" + k = 1.13850193174e-008 + nodes = self.optimization.nVertices() + t = k * nodes * nodes + refreshRate = int(5.0 / t) + if refreshRate < 1: refreshRate = 1; + if refreshRate > 1500: refreshRate = 1500; + #print "refreshRate: " + str(refreshRate) + + tolerance = 5 + initTemp = 100 + centerNdx = 0 + + selection = self.networkCanvas.getSelection() + if len(selection) > 0: + centerNdx = selection[0] + + #print "center ndx: " + str(centerNdx) + initTemp = self.optimization.radialFruchtermanReingold(centerNdx, refreshRate, initTemp) + self.networkCanvas.circles = [10000 / 12, 10000/12*2, 10000/12*3]#, 10000/12*4, 10000/12*5] + #self.networkCanvas.circles = [100, 200, 300] + self.updateCanvas() + self.networkCanvas.circles = [] + + def circularOriginal(self): + #print "Circular Original" + if self.optimization != None: + self.optimization.circularOriginal() + self.updateCanvas() + + def circularRandom(self): + #print "Circular Random" + if self.optimization != None: + self.optimization.circularRandom() + self.updateCanvas() + + def circularCrossingReduction(self): + #print "Circular Crossing Reduction" + if self.optimization != None: + self.optimization.circularCrossingReduction() + self.updateCanvas() + + def pivotMDS(self): + if self.vertexDistance is None: + self.information('Set distance matrix to input signal') + return + + if self.optimization is None: + self.information('No network found') + return + + if self.vertexDistance.dim != self.optimization.graph.nVertices: + self.error('Distance matrix dimensionality must equal number of vertices') + return + + self.frSteps = min(self.frSteps, self.vertexDistance.dim) + qApp.processEvents() + mds = orngMDS.PivotMDS(self.vertexDistance, self.frSteps) + x,y = mds.optimize() + self.optimization.graph.coors[0] = x + self.optimization.graph.coors[1] = y + self.updateCanvas() + + + """ + Network Visualization + """ + + def clickedAttLstBox(self): + if self.optimization is None: + return + + self.lastLabelColumns = set([self.attributes[i][0] for i in self.markerAttributes]) + self.networkCanvas.setLabelText(self.lastLabelColumns) + self.networkCanvas.updateData() + self.networkCanvas.replot() + + def clickedTooltipLstBox(self): + if self.optimization is None: + return + + self.lastTooltipColumns = set([self.attributes[i][0] for i in self.tooltipAttributes]) + self.networkCanvas.setTooltipText(self.lastTooltipColumns) + self.networkCanvas.updateData() + self.networkCanvas.replot() + + def clickedEdgeLabelListBox(self): + if self.optimization is None: + return + + self.lastEdgeLabelAttributes = set([self.edgeAttributes[i][0] for i in self.edgeLabelAttributes]) + self.networkCanvas.setEdgeLabelText(self.lastEdgeLabelAttributes) + self.networkCanvas.updateData() + self.networkCanvas.replot() + + def setVertexColor(self): + if self.optimization is None: + return + + self.networkCanvas.setVertexColor(self.colorCombo.currentText()) + self.lastColorColumn = self.colorCombo.currentText() + self.networkCanvas.updateData() + self.networkCanvas.replot() + + def setEdgeColor(self): + if self.optimization is None: + return + + self.networkCanvas.setEdgeColor(self.edgeColorCombo.currentText()) + self.lastEdgeColorColumn = self.edgeColorCombo.currentText() + self.networkCanvas.updateData() + self.networkCanvas.replot() + + def setGraphGrid(self): + self.networkCanvas.enableGridY(self.networkCanvasShowGrid) + self.networkCanvas.enableGridX(self.networkCanvasShowGrid) + + def selectAllConnectedNodes(self): + self.networkCanvas.selectConnectedNodes(1000000) + + def setShowMissingValues(self): + self.networkCanvas.showMissingValues = self.showMissingValues + self.networkCanvas.updateData() + self.networkCanvas.replot() + + def setMaxLinkSize(self): + if self.optimization is None: + return + + self.networkCanvas.maxEdgeSize = self.maxLinkSize + self.networkCanvas.setEdgesSize() + self.networkCanvas.replot() + + def setVertexSize(self): + if self.optimization is None or self.networkCanvas is None: + return + + if self.minVertexSize > self.maxVertexSize: + self.maxVertexSize = self.minVertexSize + + self.networkCanvas.minVertexSize = self.minVertexSize + self.networkCanvas.maxVertexSize = self.maxVertexSize + self.lastVertexSizeColumn = self.vertexSizeCombo.currentText() + + if self.vertexSize > 0: + self.networkCanvas.setVerticesSize(self.lastVertexSizeColumn, self.invertSize) + else: + self.networkCanvas.setVerticesSize() + + self.networkCanvas.replot() + + def setFontSize(self): + if self.networkCanvas is None: + return + + self.networkCanvas.fontSize = self.fontSize + self.networkCanvas.drawPlotItems() + + def setRenderAntialiased(self): + self.networkCanvas.renderAntialiased = self.renderAntialiased + self.networkCanvas.updateData() + self.networkCanvas.replot() + + def sendReport(self): + self.reportSettings("Graph data", + [("Number of vertices", self.nVertices), + ("Number of edges", self.nEdges), + ("Vertices per edge", "%.3f" % self.verticesPerEdge), + ("Edges per vertex", "%.3f" % self.edgesPerVertex), + ("Diameter", self.diameter), + ("Clustering Coefficient", "%.1f%%" % self.clustering_coefficient) + ]) + if self.color or self.vertexSize or self.markerAttributes or self.edgeColor: + self.reportSettings("Visual settings", + [self.color and ("Vertex color", self.colorCombo.currentText()), + self.vertexSize and ("Vertex size", str(self.vertexSizeCombo.currentText()) + " (inverted)" if self.invertSize else ""), + self.markerAttributes and ("Labels", ", ".join(self.attributes[i][0] for i in self.markerAttributes)), + self.edgeColor and ("Edge colors", self.edgeColorCombo.currentText()), + ]) + self.reportSettings("Optimization", + [("Method", self.optCombo.currentText()), + ("Iterations", self.frSteps)]) + self.reportSection("Graph") + self.reportImage(self.networkCanvas.saveToFileDirect) + + +if __name__=="__main__": + appl = QApplication(sys.argv) + ow = OWNetExplorer() + ow.show() + appl.exec_() + Index: Orange/OrangeWidgets/Unsupervised/OWNetworkCanvas.py =================================================================== --- Orange/OrangeWidgets/Unsupervised/OWNetworkCanvas.py (revision 9671) +++ Orange/OrangeWidgets/Unsupervised/OWNetworkCanvas.py (revision 9671) @@ -0,0 +1,1280 @@ +CIRCLE = 0 +SQUARE = 1 +ROUND_RECT = 2 + +NOTHING = 0 +ZOOMING = 1 +SELECT_RECTANGLE = 2 +SELECT_POLYGON = 3 +MOVE_SELECTION = 100 + +from OWGraph import * +from numpy import * +from orngNetwork import Network +from orngScaleScatterPlotData import * + +class NetworkVertex(): + def __init__(self): + self.index = -1 + self.marked = False + self.show = True + self.highlight = False + self.selected = False + self.label = [] + self.tooltip = [] + self.uuid = None + + self.image = None + self.pen = QPen(Qt.blue, 1) + self.pen.setJoinStyle(Qt.RoundJoin) + self.nocolor = Qt.white + self.color = Qt.blue + self.size = 5 + self.style = 1 + +class NetworkEdge(): + def __init__(self): + self.u = None + self.v = None + self.links_index = None + self.arrowu = 0 + self.arrowv = 0 + self.weight = 0 + self.label = [] + + self.pen = QPen(Qt.lightGray, 1) + self.pen.setCapStyle(Qt.RoundCap) + +class NetworkCurve(QwtPlotCurve): + def __init__(self, parent, pen=QPen(Qt.black), xData=None, yData=None): + QwtPlotCurve.__init__(self, "Network Curve") + + self.coors = None + self.vertices = [] + self.edges = [] + self.setItemAttribute(QwtPlotItem.Legend, 0) + self.showEdgeLabels = 0 + + def moveSelectedVertices(self, dx, dy): + selected = self.getSelectedVertices() + + self.coors[0][selected] = self.coors[0][selected] + dx + self.coors[1][selected] = self.coors[1][selected] + dy + + self.setData(self.coors[0], self.coors[1]) + return selected + + def setVertexColor(self, v, color): + pen = self.vertices[v].pen + self.vertices[v].color = color + self.vertices[v].pen = QPen(color, pen.width()) + + def setEdgeColor(self, index, color, nocolor=0): + pen = self.edges[index].pen + if nocolor: + color.setAlpha(0) + self.edges[index].pen = QPen(color, pen.width()) + self.edges[index].pen.setCapStyle(Qt.RoundCap) + + def getSelectedVertices(self): + return [vertex.index for vertex in self.vertices if vertex.selected] + + def getUnselectedVertices(self): + return [vertex.index for vertex in self.vertices if not vertex.selected] + + def getMarkedVertices(self): + return [vertex.index for vertex in self.vertices if vertex.marked] + + def setMarkedVertices(self, vertices): + for vertex in self.vertices: + if vertex.index in vertices: + vertex.marked = True + else: + vertex.marked = False + + def markToSel(self): + for vertex in self.vertices: + if vertex.marked == True: + vertex.selected = True + + def selToMark(self): + for vertex in self.vertices: + if vertex.selected == True: + vertex.selected = False + vertex.marked = True + + def unMark(self): + for vertex in self.vertices: + vertex.marked = False + + def unSelect(self): + for vertex in self.vertices: + vertex.selected = False + + def setHiddenVertices(self, nodes): + for vertex in self.vertices: + if vertex.index in nodes: + vertex.show = False + else: + vertex.show = True + + def hideSelectedVertices(self): + for vertex in self.vertices: + if vertex.selected: + vertex.show = False + + def hideUnSelectedVertices(self): + for vertex in self.vertices: + if not vertex.selected: + vertex.show = False + + def showAllVertices(self): + for vertex in self.vertices: + vertex.show = True + + def changed(self): + self.itemChanged() + + def draw(self, painter, xMap, yMap, rect): + for edge in self.edges: + if edge.u.show and edge.v.show: + painter.setPen(edge.pen) + + px1 = xMap.transform(self.coors[0][edge.u.index]) #ali pa tudi self.x1, itd + py1 = yMap.transform(self.coors[1][edge.u.index]) + px2 = xMap.transform(self.coors[0][edge.v.index]) + py2 = yMap.transform(self.coors[1][edge.v.index]) + + painter.drawLine(px1, py1, px2, py2) + + d = 12 + #painter.setPen(QPen(Qt.lightGray, 1)) + painter.setBrush(Qt.lightGray) + if edge.arrowu: + x = px1 - px2 + y = py1 - py2 + + fi = math.atan2(y, x) * 180 * 16 / math.pi + + if not fi is None: + # (180*16) - fi - (20*16), (40*16) + painter.drawPie(px1 - d, py1 - d, 2 * d, 2 * d, 2560 - fi, 640) + + if edge.arrowv: + x = px1 - px2 + y = py1 - py2 + + fi = math.atan2(y, x) * 180 * 16 / math.pi + if not fi is None: + # (180*16) - fi - (20*16), (40*16) + painter.drawPie(px1 - d, py1 - d, 2 * d, 2 * d, 2560 - fi, 640) + + if self.showEdgeLabels and len(edge.label) > 0: + lbl = ', '.join(edge.label) + x = (px1 + px2) / 2 + y = (py1 + py2) / 2 + + th = painter.fontMetrics().height() + tw = painter.fontMetrics().width(lbl) + r = QRect(x - tw / 2, y - th / 2, tw, th) + painter.fillRect(r, QBrush(Qt.white)) + painter.drawText(r, Qt.AlignHCenter + Qt.AlignVCenter, lbl) + + for vertex in self.vertices: + if vertex.show: + pX = xMap.transform(self.coors[0][vertex.index]) #dobimo koordinati v pikslih (tipa integer) + pY = yMap.transform(self.coors[1][vertex.index]) #ki se stejeta od zgornjega levega kota canvasa + if vertex.selected: + painter.setPen(QPen(Qt.yellow, 3)) + painter.setBrush(vertex.color) + painter.drawEllipse(pX - (vertex.size + 4) / 2, pY - (vertex.size + 4) / 2, vertex.size + 4, vertex.size + 4) + elif vertex.marked: + painter.setPen(vertex.pen) + painter.setBrush(vertex.color) + painter.drawEllipse(pX - vertex.size / 2, pY - vertex.size / 2, vertex.size, vertex.size) + else: + painter.setPen(vertex.pen) + painter.setBrush(vertex.nocolor) + #print pX - vertex.size / 2, pY - vertex.size / 2, vertex.size + painter.drawEllipse(pX - vertex.size / 2, pY - vertex.size / 2, vertex.size, vertex.size) + +class OWNetworkCanvas(OWGraph): + def __init__(self, master, parent=None, name="None"): + OWGraph.__init__(self, parent, name) + self.master = master + self.parent = parent + self.labelText = [] + self.tooltipText = [] + self.vertices_old = {} # distionary of nodes (orngIndex: vertex_objekt) + self.edges_old = {} # distionary of edges (curveKey: edge_objekt) + self.vertices = [] + self.edges = [] + self.indexPairs = {} # distionary of type CurveKey: orngIndex (for nodes) + #self.selection = [] # list of selected nodes (indices) + self.markerKeys = {} # dictionary of type NodeNdx : markerCurveKey + self.tooltipKeys = {} # dictionary of type NodeNdx : tooltipCurveKey + self.visualizer = None + self.vertexDegree = [] # seznam vozlisc oblike (vozlisce, stevilo povezav), sortiran po stevilu povezav + self.edgesKey = -1 + #self.vertexSize = 6 + self.nVertices = 0 + self.enableXaxis(0) + self.enableYLaxis(0) + self.state = NOTHING #default je rocno premikanje + self.hiddenNodes = [] + self.markedNodes = set() + self.markWithRed = False + self.circles = [] + self.tooltipNeighbours = 2 + self.selectionNeighbours = 2 + self.freezeNeighbours = False + self.labelsOnMarkedOnly = 0 + self.enableWheelZoom = 1 + self.optimizing = 0 + self.stopOptimizing = 0 + self.insideview = 0 + self.insideviewNeighbours = 2 + self.enableGridXB(False) + self.enableGridYL(False) + self.renderAntialiased = 1 + self.sendMarkedNodes = None + self.showEdgeLabels = 0 + self.showDistances = 0 + self.showMissingValues = 0 + + self.showWeights = 0 + self.showIndexes = 0 + self.minEdgeWeight = sys.maxint + self.maxEdgeWeight = 0 + self.maxEdgeSize = 1 + + self.maxVertexSize = 5 + self.minVertexSize = 5 + self.invertEdgeSize = 0 + self.showComponentAttribute = None + self.forceVectors = None + self.appendToSelection = 1 + self.fontSize = 12 + + self.setAxisAutoScale(self.xBottom) + self.setAxisAutoScale(self.yLeft) + + self.networkCurve = NetworkCurve(self) + self.callbackMoveVertex = None + self.callbackSelectVertex = None + self.minComponentEdgeWidth = 0 + self.maxComponentEdgeWidth = 0 + self.vertexDistance = None + self.controlPressed = False + self.altPressed = False + + self.setFocusPolicy(Qt.StrongFocus) + + def getSelection(self): + return self.networkCurve.getSelectedVertices() + + def getMarkedVertices(self): + return self.networkCurve.getMarkedVertices() + + def getVertexSize(self, index): + return 6 + + def setHiddenVertices(self, nodes): + self.networkCurve.setHiddenVertices(nodes) + + def hideSelectedVertices(self): + self.networkCurve.hideSelectedVertices() + self.drawPlotItems() + + def hideUnSelectedVertices(self): + self.networkCurve.hideUnSelectedVertices() + self.drawPlotItems() + + def showAllVertices(self): + self.networkCurve.showAllVertices() + self.drawPlotItems() + + def optimize(self, frSteps): + qApp.processEvents() + tolerance = 5 + initTemp = 100 + breakpoints = 20 + k = int(frSteps / breakpoints) + o = frSteps % breakpoints + iteration = 0 + coolFactor = exp(log(10.0 / 10000.0) / frSteps) + #print coolFactor + if k > 0: + while iteration < breakpoints: + initTemp = self.visualizer.fruchtermanReingold(k, initTemp, coolFactor, self.hiddenNodes) + iteration += 1 + qApp.processEvents() + self.updateCanvas() + + initTemp = self.visualizer.fruchtermanReingold(o, initTemp, coolFactor, self.hiddenNodes) + qApp.processEvents() + self.updateCanvas() + else: + while iteration < o: + initTemp = self.visualizer.fruchtermanReingold(1, initTemp, coolFactor, self.hiddenNodes) + iteration += 1 + qApp.processEvents() + self.updateCanvas() + + def markedToSelection(self): + self.networkCurve.markToSel() + self.drawPlotItems() + + def selectionToMarked(self): + self.networkCurve.selToMark() + self.drawPlotItems() + + if self.sendMarkedNodes != None: + self.sendMarkedNodes(self.networkCurve.getMarkedVertices()) + + def removeSelection(self, replot=True): + self.networkCurve.unSelect() + + if replot: + self.replot() + + def selectNeighbours(self, sel, nodes, depth, maxdepth): + #print "list: " + str(sel) + #print "nodes: " + str(nodes) + sel.update(nodes) + if depth < maxdepth: + for i in nodes: + neighbours = set(self.visualizer.graph.getNeighbours(i)) + #print "neighbours: " + str(neighbours) + self.selectNeighbours(sel, neighbours - sel, depth + 1, maxdepth) + + def getSelectedExamples(self): + selection = self.networkCurve.getSelectedVertices() + + if len(selection) == 0: + return None + + if self.visualizer.graph.items != None: + return self.visualizer.graph.items.getitems(selection) + else: + return None + + def getUnselectedExamples(self): + unselection = self.networkCurve.getUnselectedVertices() + + if len(unselection) == 0: + return None + + if self.visualizer.graph.items != None: + return self.visualizer.graph.items.getitems(unselection) + else: + return None + + def getSelectedGraph(self): + selection = self.networkCurve.getSelectedVertices() + + if len(selection) == 0: + return None + + print self.visualizer.graph + subgraph = self.visualizer.graph.getSubGraph(selection) + subnet = Network(subgraph) + return subnet + + def getSelectedVertices(self): + return self.networkCurve.getSelectedVertices() + + def getNeighboursUpTo(self, ndx, dist): + newNeighbours = neighbours = set([ndx]) + for d in range(dist): + tNewNeighbours = set() + for v in newNeighbours: + tNewNeighbours |= set(self.visualizer.graph.getNeighbours(v)) + newNeighbours = tNewNeighbours - neighbours + neighbours |= newNeighbours + return neighbours + + def markSelectionNeighbours(self): + if not self.freezeNeighbours and self.selectionNeighbours: + toMark = set() + for ndx in self.networkCurve.getSelectedVertices(): + toMark |= self.getNeighboursUpTo(ndx, self.selectionNeighbours) + + self.networkCurve.setMarkedVertices(toMark) + self.drawPlotItems() + + elif not self.freezeNeighbours and self.selectionNeighbours == 0: + self.networkCurve.setMarkedVertices(self.networkCurve.getSelectedVertices()) + self.drawPlotItems() + + if self.sendMarkedNodes != None: + self.sendMarkedNodes(self.networkCurve.getMarkedVertices()) + + def unMark(self): + self.networkCurve.unMark() + self.drawPlotItems(replot=0) + + if self.sendMarkedNodes != None: + self.sendMarkedNodes([]) + + def setMarkedVertices(self, vertices): + self.networkCurve.setMarkedVertices(vertices) + self.drawPlotItems(replot=0) + + if self.sendMarkedNodes != None: + self.sendMarkedNodes(self.networkCurve.getMarkedVertices()) + + def activateMoveSelection(self): + self.state = MOVE_SELECTION + + def mouseMoveEvent(self, event): + if not self.visualizer: + return + + if self.mouseCurrentlyPressed and self.state == MOVE_SELECTION and self.GMmouseMoveEvent != None: + newX = self.invTransform(2, event.pos().x()) + newY = self.invTransform(0, event.pos().y()) + + dx = newX - self.invTransform(2, self.GMmouseMoveEvent.x()) + dy = newY - self.invTransform(0, self.GMmouseMoveEvent.y()) + movedVertices = self.networkCurve.moveSelectedVertices(dx, dy) + + self.GMmouseMoveEvent.setX(event.pos().x()) #zacetni dogodek postane trenutni + self.GMmouseMoveEvent.setY(event.pos().y()) + + self.drawPlotItems(replot=1, vertices=movedVertices) + if self.callbackMoveVertex: + self.callbackMoveVertex() + else: + OWGraph.mouseMoveEvent(self, event) + + if not self.freezeNeighbours and self.tooltipNeighbours: + px = self.invTransform(2, event.x()) + py = self.invTransform(0, event.y()) + ndx, mind = self.visualizer.closestVertex(px, py) + dX = self.transform(QwtPlot.xBottom, self.visualizer.graph.coors[0][ndx]) - event.x() + dY = self.transform(QwtPlot.yLeft, self.visualizer.graph.coors[1][ndx]) - event.y() + # transform to pixel distance + distance = math.sqrt(dX**2 + dY**2) + + if ndx != -1 and distance <= self.vertices[ndx].size / 2: + toMark = set(self.getNeighboursUpTo(ndx, self.tooltipNeighbours)) + self.networkCurve.setMarkedVertices(toMark) + self.drawPlotItems() + + if self.sendMarkedNodes != None: + self.sendMarkedNodes(self.networkCurve.getMarkedVertices()) + else: + self.networkCurve.unMark() + self.drawPlotItems() + + if self.sendMarkedNodes != None: + self.sendMarkedNodes([]) + + if self.showDistances: + selection = self.networkCurve.getSelectedVertices() + if len(selection) > 0: + px = self.invTransform(2, event.x()) + py = self.invTransform(0, event.y()) + + v, mind = self.visualizer.closestVertex(px, py) + dX = self.transform(QwtPlot.xBottom, self.visualizer.graph.coors[0][v]) - event.x() + dY = self.transform(QwtPlot.yLeft, self.visualizer.graph.coors[1][v]) - event.y() + # transform to pixel distance + distance = math.sqrt(dX**2 + dY**2) + if v != -1 and distance <= self.vertices[v].size / 2: + if self.visualizer.vertexDistance == None: + dst = 'vertex distance signal not set' + else: + dst = 0 + for u in selection: + dst += self.visualizer.vertexDistance[u, v] + dst = dst / len(selection) + + self.showTip(event.pos().x(), event.pos().y(), str(dst)) + self.replot() + + def mousePressEvent(self, event): + if not self.visualizer: + return + + #self.grabKeyboard() + self.mouseSelectedVertex = 0 + self.GMmouseMoveEvent = None + + if self.state == MOVE_SELECTION: + self.mouseCurrentlyPressed = 1 + #if self.isPointSelected(self.invTransform(self.xBottom, event.pos().x()), self.invTransform(self.yLeft, event.pos().y())) and self.selection != []: + # self.GMmouseStartEvent = QPoint(event.pos().x(), event.pos().y()) + #else: + # button pressed outside selected area or there is no area + self.selectVertex(event.pos()) + self.GMmouseStartEvent = QPoint(event.pos().x(), event.pos().y()) + self.replot() + elif self.state == SELECT_RECTANGLE: + self.GMmouseStartEvent = QPoint(event.pos().x(), event.pos().y()) + + if self.clickedSelectedOnVertex(event.pos()): + self.mouseSelectedVertex = 1 + self.mouseCurrentlyPressed = 1 + self.state = MOVE_SELECTION + self.GMmouseMoveEvent = QPoint(event.pos().x(), event.pos().y()) + elif self.clickedOnVertex(event.pos()): + self.mouseSelectedVertex = 1 + self.mouseCurrentlyPressed = 1 + else: + OWGraph.mousePressEvent(self, event) + else: + OWGraph.mousePressEvent(self, event) + + def mouseReleaseEvent(self, event): + if not self.visualizer: + return + + #self.releaseKeyboard() + if self.state == MOVE_SELECTION: + self.state = SELECT_RECTANGLE + self.mouseCurrentlyPressed = 0 + + self.moveGroup = False + #self.GMmouseStartEvent=None + + if self.state == SELECT_RECTANGLE: + x1 = self.invTransform(2, self.GMmouseStartEvent.x()) + y1 = self.invTransform(0, self.GMmouseStartEvent.y()) + + x2 = self.invTransform(2, event.pos().x()) + y2 = self.invTransform(0, event.pos().y()) + + + if self.mouseSelectedVertex == 1 and x1 == x2 and y1 == y2 and self.selectVertex(self.GMmouseStartEvent): + QwtPlot.mouseReleaseEvent(self, event) + elif self.mouseSelectedVertex == 0: + + selection = self.visualizer.getVerticesInRect(x1, y1, x2, y2) + + def selectVertex(ndx): + if self.vertices[ndx].show: + self.vertices[ndx].selected = True + + map(selectVertex, selection) + + if len(selection) == 0 and x1 == x2 and y1 == y2: + self.removeSelection() + self.unMark() + + self.markSelectionNeighbours() + OWGraph.mouseReleaseEvent(self, event) + self.removeAllSelections() + + elif self.state == SELECT_POLYGON: + OWGraph.mouseReleaseEvent(self, event) + if self.tempSelectionCurve == None: #if OWVisGraph closed polygon + self.selectVertices() + else: + OWGraph.mouseReleaseEvent(self, event) + + self.mouseCurrentlyPressed = 0 + self.moveGroup = False + + if self.callbackSelectVertex != None: + self.callbackSelectVertex() + + def keyPressEvent(self, e): + if not self.visualizer: + return + + if e.key() == 87 or e.key() == 81: + selection = [v.index for v in self.vertices if v.selected] + if len(selection) > 0: + phi = [math.pi / -180 if e.key() == 87 else math.pi / 180] + self.visualizer.rotateVertices([selection], phi) + self.drawPlotItems(replot=1) + + if e.key() == Qt.Key_Control: + self.controlPressed = True + + elif e.key() == Qt.Key_Alt: + self.altPressed = True + + if e.text() == "f": + self.graph.freezeNeighbours = not self.graph.freezeNeighbours + + OWGraph.keyPressEvent(self, e) + + def keyReleaseEvent(self, e): + if e.key() == Qt.Key_Control: + self.controlPressed = False + + elif e.key() == Qt.Key_Alt: + self.altPressed = False + + OWGraph.keyReleaseEvent(self, e) + + + def clickedSelectedOnVertex(self, pos): + min = 1000000 + ndx = -1 + + px = self.invTransform(2, pos.x()) + py = self.invTransform(0, pos.y()) + + ndx, min = self.visualizer.closestVertex(px, py) + dX = self.transform(QwtPlot.xBottom, self.visualizer.graph.coors[0][ndx]) - pos.x() + dY = self.transform(QwtPlot.yLeft, self.visualizer.graph.coors[1][ndx]) - pos.y() + # transform to pixel distance + distance = math.sqrt(dX**2 + dY**2) + if ndx != -1 and distance <= self.vertices[ndx].size / 2: + return self.vertices[ndx].selected + else: + return False + + def clickedOnVertex(self, pos): + min = 1000000 + ndx = -1 + + px = self.invTransform(2, pos.x()) + py = self.invTransform(0, pos.y()) + + ndx, min = self.visualizer.closestVertex(px, py) + dX = self.transform(QwtPlot.xBottom, self.visualizer.graph.coors[0][ndx]) - pos.x() + dY = self.transform(QwtPlot.yLeft, self.visualizer.graph.coors[1][ndx]) - pos.y() + # transform to pixel distance + distance = math.sqrt(dX**2 + dY**2) + if ndx != -1 and distance <= self.vertices[ndx].size / 2: + return True + else: + return False + + def selectVertex(self, pos): + min = 1000000 + ndx = -1 + + px = self.invTransform(2, pos.x()) + py = self.invTransform(0, pos.y()) + + ndx, min = self.visualizer.closestVertex(px, py) + + dX = self.transform(QwtPlot.xBottom, self.visualizer.graph.coors[0][ndx]) - pos.x() + dY = self.transform(QwtPlot.yLeft, self.visualizer.graph.coors[1][ndx]) - pos.y() + # transform to pixel distance + distance = math.sqrt(dX**2 + dY**2) + if ndx != -1 and distance <= self.vertices[ndx].size / 2: + if not self.appendToSelection and not self.controlPressed: + self.removeSelection() + + if self.insideview: + self.networkCurve.unSelect() + self.vertices[ndx].selected = not self.vertices[ndx].selected + self.optimize(100) + + self.markSelectionNeighbours() + else: + self.vertices[ndx].selected = not self.vertices[ndx].selected + self.markSelectionNeighbours() + + return True + else: + return False + self.removeSelection() + self.unMark() + + def updateData(self): + if self.visualizer == None: + return + + self.removeDrawingCurves(removeLegendItems=0) + self.tips.removeAll() + + if self.vertexDistance and self.minComponentEdgeWidth > 0 and self.maxComponentEdgeWidth > 0: + components = self.visualizer.graph.getConnectedComponents() + matrix = self.vertexDistance.avgLinkage(components) + + edges = set() + for u in range(matrix.dim): + neighbours = matrix.getKNN(u, 2) + for v in neighbours: + if v < u: + edges.add((v, u)) + else: + edges.add((u, v)) + edges = list(edges) +# show 2n strongest edges +# vals = matrix.getValues() +# vals = zip(vals, range(len(vals))) +# count = 0 +# add = 0 +# for i in range(matrix.dim): +# add += i + 1 +# for j in range(i+1, matrix.dim): +# v, ind = vals[count] +# ind += add +# vals[count] = (v, ind) +# count += 1 +# vals.sort(reverse=0) +# vals = vals[:2 * matrix.dim] +# edges = [(ind / matrix.dim, ind % matrix.dim) for v, ind in vals] +# print "number of component edges:", len(edges), "number of components:", len(components) + components_c = [(sum(self.visualizer.graph.coors[0][c]) / len(c), sum(self.visualizer.graph.coors[1][c]) / len(c)) for c in components] + weights = [1 - matrix[u,v] for u,v in edges] + + max_weight = max(weights) + min_weight = min(weights) + span_weights = max_weight - min_weight + + for u,v in edges: + x = [components_c[u][0], components_c[v][0]] + y = [components_c[u][1], components_c[v][1]] + w = ((1 - matrix[u,v]) - min_weight) / span_weights * (self.maxComponentEdgeWidth - self.minComponentEdgeWidth) + self.minComponentEdgeWidth + + pen = QPen() + pen.setWidth(w) + pen.setBrush(QColor(50,200,255,15)) + pen.setCapStyle(Qt.FlatCap) + pen.setJoinStyle(Qt.RoundJoin) + self.addCurve("component_edges", Qt.green, Qt.green, 0, style=QwtPlotCurve.Lines, symbol = QwtSymbol.NoSymbol, xData=x, yData=y, pen=pen, showFilledSymbols=False) + + + self.networkCurve.setData(self.visualizer.network.coors[0], self.visualizer.network.coors[1]) + + if self.insideview == 1: + selection = self.networkCurve.getSelectedVertices() + if len(selection) >= 1: + visible = set() + visible |= set(selection) + visible |= self.getNeighboursUpTo(selection[0], self.insideviewNeighbours) + self.networkCurve.setHiddenVertices(set(range(self.nVertices)) - visible) + + edgesCount = 0 + + if self.forceVectors != None: + for v in self.forceVectors: + self.addCurve("force", Qt.white, Qt.green, 1, style=QwtPlotCurve.Lines, xData=v[0], yData=v[1], showFilledSymbols=False) + + for r in self.circles: + step = 2 * pi / 64; + fi = 0 + x = [] + y = [] + for i in range(65): + x.append(r * cos(fi) + 5000) + y.append(r * sin(fi) + 5000) + fi += step + + self.addCurve("radius", Qt.white, Qt.green, 1, style=QwtPlotCurve.Lines, xData=x, yData=y, showFilledSymbols=False) + + if self.renderAntialiased: + self.networkCurve.setRenderHint(QwtPlotItem.RenderAntialiased) + else: + self.networkCurve.setRenderHint(QwtPlotItem.RenderAntialiased, False) + + self.networkCurve.showEdgeLabels = self.showEdgeLabels + self.networkCurve.attach(self) + self.drawPlotItems(replot=0) + + #self.zoomExtent() + + def drawPlotItems(self, replot=1, vertices=[]): + if len(vertices) > 0: + for vertex in vertices: + x1 = float(self.visualizer.network.coors[0][vertex]) + y1 = float(self.visualizer.network.coors[1][vertex]) + + if vertex in self.markerKeys: + mkey = self.markerKeys[vertex] + mkey.setValue(x1, y1) + + if 'index ' + str(vertex) in self.markerKeys: + mkey = self.markerKeys['index ' + str(vertex)] + mkey.setValue(x1, y1) + + if vertex in self.tooltipKeys: + tkey = self.tooltipKeys[vertex] + self.tips.positions[tkey] = (x1, y1, 0, 0) + else: + self.markerKeys = {} + self.removeMarkers() + self.drawLabels() + self.drawToolTips() + self.drawWeights() + self.drawIndexes() + self.drawComponentKeywords() + + if replot: + self.replot() + + def drawComponentKeywords(self): + if self.showComponentAttribute == None: + return + + if self.visualizer == None or self.visualizer.graph == None or self.visualizer.graph.items == None: + return + + if str(self.showComponentAttribute) not in self.visualizer.graph.items.domain: + self.showComponentAttribute = None + return + + components = self.visualizer.graph.getConnectedComponents() + + for component in components: + if len(component) == 0: + continue + + vertices = [vertex for vertex in component if self.vertices[vertex].show] + + if len(vertices) == 0: + continue + + xes = [self.visualizer.network.coors[0][vertex] for vertex in vertices] + yes = [self.visualizer.network.coors[1][vertex] for vertex in vertices] + + x1 = sum(xes) / len(xes) + y1 = sum(yes) / len(yes) + + lbl = str(self.visualizer.graph.items[component[0]][str(self.showComponentAttribute)]) + + mkey = self.addMarker(lbl, float(x1), float(y1), alignment=Qt.AlignCenter, size=self.fontSize) + + def drawToolTips(self): + # add ToolTips + self.tooltipData = [] + self.tooltipKeys = {} + self.tips.removeAll() + if len(self.tooltipText) > 0: + for vertex in self.vertices: + if not vertex.show: + continue + + x1 = self.visualizer.network.coors[0][vertex.index] + y1 = self.visualizer.network.coors[1][vertex.index] + lbl = "" + values = self.visualizer.graph.items[vertex.index] + for ndx in self.tooltipText: + if not ndx in self.visualizer.graph.items.domain: + continue + + value = str(values[ndx]) + # wrap text + i = 0 + while i < len(value) / 100: + value = value[:((i + 1) * 100) + i] + "\n" + value[((i + 1) * 100) + i:] + i += 1 + + lbl = lbl + str(value) + "\n" + + if lbl != '': + lbl = lbl[:-1] + self.tips.addToolTip(x1, y1, lbl) + self.tooltipKeys[vertex.index] = len(self.tips.texts) - 1 + + def drawLabels(self): + if len(self.labelText) > 0: + for vertex in self.vertices: + if not vertex.show: + continue + + if self.labelsOnMarkedOnly and not (vertex.marked): + continue + + x1 = self.visualizer.network.coors[0][vertex.index] + y1 = self.visualizer.network.coors[1][vertex.index] + lbl = "" + values = self.visualizer.graph.items[vertex.index] + if self.showMissingValues: + lbl = ", ".join([str(values[ndx]) for ndx in self.labelText]) + else: + lbl = ", ".join([str(values[ndx]) for ndx in self.labelText if str(values[ndx]) != '?']) + #if not self.showMissingValues and lbl == '': + # continue + + if lbl: + vertex.label = lbl + mkey = self.addMarker(lbl, float(x1), float(y1), alignment=Qt.AlignBottom, size=self.fontSize) + self.markerKeys[vertex.index] = mkey + + def drawIndexes(self): + if self.showIndexes: + for vertex in self.vertices: + if not vertex.show: + continue + + if self.labelsOnMarkedOnly and not (vertex.marked): + continue + + x1 = self.visualizer.network.coors[0][vertex.index] + y1 = self.visualizer.network.coors[1][vertex.index] + + lbl = str(vertex.index) + mkey = self.addMarker(lbl, float(x1), float(y1), alignment=Qt.AlignTop, size=self.fontSize) + self.markerKeys['index ' + str(vertex.index)] = mkey + + def drawWeights(self): + if self.showWeights: + for edge in self.edges: + if not (edge.u.show and edge.v.show): + continue + + if self.labelsOnMarkedOnly and not (edge.u.marked and edge.v.marked): + continue + + x1 = (self.visualizer.network.coors[0][edge.u.index] + self.visualizer.network.coors[0][edge.v.index]) / 2 + y1 = (self.visualizer.network.coors[1][edge.u.index] + self.visualizer.network.coors[1][edge.v.index]) / 2 + + if edge.weight == None: + lbl = "None" + else: + lbl = "%.2f" % float(edge.weight) + + mkey = self.addMarker(lbl, float(x1), float(y1), alignment=Qt.AlignCenter, size=self.fontSize) + self.markerKeys[(edge.u, edge.v)] = mkey + + def getColorIndeces(self, table, attribute, palette): + colorIndices = {} + colorIndex = None + minValue = None + maxValue = None + + if attribute[0] != "(" or attribute[ -1] != ")": + i = 0 + for var in table.domain.variables: + if var.name == attribute: + colorIndex = i + if var.varType == orange.VarTypes.Discrete: + colorIndices = getVariableValueIndices(var, colorIndex) + + i += 1 + metas = table.domain.getmetas() + for i, var in metas.iteritems(): + if var.name == attribute: + colorIndex = i + if var.varType == orange.VarTypes.Discrete: + colorIndices = getVariableValueIndices(var, colorIndex) + + colorIndices['?'] = len(colorIndices) + palette.setNumberOfColors(len(colorIndices)) + + if colorIndex != None and table.domain[colorIndex].varType == orange.VarTypes.Continuous: + minValue = float(min([x[colorIndex].value for x in table if x[colorIndex].value != "?"] or [0.0])) + maxValue = float(max([x[colorIndex].value for x in table if x[colorIndex].value != "?"] or [0.0])) + + return colorIndices, colorIndex, minValue, maxValue + + def setEdgeColor(self, attribute): + if self.visualizer == None: + return + + colorIndices, colorIndex, minValue, maxValue = self.getColorIndeces(self.visualizer.graph.links, attribute, self.discEdgePalette) + + for index in range(self.nEdges): + if colorIndex != None: + links_index = self.networkCurve.edges[index].links_index + if links_index == None: + continue + + if self.visualizer.graph.links.domain[colorIndex].varType == orange.VarTypes.Continuous: + newColor = self.discEdgePalette[0] + if str(self.visualizer.graph.links[links_index][colorIndex]) != "?": + if maxValue == minValue: + newColor = self.discEdgePalette[0] + else: + value = (float(self.visualizer.graph.links[links_index][colorIndex].value) - minValue) / (maxValue - minValue) + newColor = self.contEdgePalette[value] + + self.networkCurve.setEdgeColor(index, newColor) + + elif self.visualizer.graph.links.domain[colorIndex].varType == orange.VarTypes.Discrete: + newColor = self.discEdgePalette[colorIndices[self.visualizer.graph.links[links_index][colorIndex].value]] + if self.visualizer.graph.links[links_index][colorIndex].value == "0": + self.networkCurve.setEdgeColor(index, newColor, nocolor=1) + else: + self.networkCurve.setEdgeColor(index, newColor) + + else: + newColor = self.discEdgePalette[0] + self.networkCurve.setEdgeColor(index, newColor) + + self.replot() + + def setVertexColor(self, attribute): + if self.visualizer == None: + return + + colorIndices, colorIndex, minValue, maxValue = self.getColorIndeces(self.visualizer.graph.items, attribute, self.discPalette) + + for v in range(self.nVertices): + if colorIndex != None: + if self.visualizer.graph.items.domain[colorIndex].varType == orange.VarTypes.Continuous: + newColor = self.discPalette[0] + + if str(self.visualizer.graph.items[v][colorIndex]) != "?": + if maxValue == minValue: + newColor = self.discPalette[0] + else: + value = (float(self.visualizer.graph.items[v][colorIndex].value) - minValue) / (maxValue - minValue) + newColor = self.contPalette[value] + + self.networkCurve.setVertexColor(v, newColor) + + elif self.visualizer.graph.items.domain[colorIndex].varType == orange.VarTypes.Discrete: + newColor = self.discPalette[colorIndices[self.visualizer.graph.items[v][colorIndex].value]] + #print newColor + self.networkCurve.setVertexColor(v, newColor) + + else: + newColor = self.discPalette[0] + self.networkCurve.setVertexColor(v, newColor) + + self.replot() + + def setLabelText(self, attributes): + self.labelText = [] + if self.visualizer == None or self.visualizer.graph == None or self.visualizer.graph.items == None: + return + + if isinstance(self.visualizer.graph.items, orange.ExampleTable): + data = self.visualizer.graph.items + for att in attributes: + for i in range(len(data.domain)): + if data.domain[i].name == att: + self.labelText.append(i) + + if self.visualizer.graph.items.domain.hasmeta(att): + self.labelText.append(self.visualizer.graph.items.domain.metaid(att)) + + def setTooltipText(self, attributes): + self.tooltipText = [] + if self.visualizer == None or self.visualizer.graph == None or self.visualizer.graph.items == None: + return + + if isinstance(self.visualizer.graph.items, orange.ExampleTable): + data = self.visualizer.graph.items + for att in attributes: + for i in range(len(data.domain)): + if data.domain[i].name == att: + self.tooltipText.append(i) + + if self.visualizer.graph.items.domain.hasmeta(att): + self.tooltipText.append(self.visualizer.graph.items.domain.metaid(att)) + + def setEdgeLabelText(self, attributes): + self.edgeLabelText = [] + if self.visualizer == None or self.visualizer.graph == None or self.visualizer.graph.items == None: + return + + def edgesContainsEdge(self, i, j): + for e in range(self.nEdges): + (key, iTmp, jTmp) = self.edges_old[e] + + if (iTmp == i and jTmp == j) or (iTmp == j and jTmp == i): + return True + return False + + def addVisualizer(self, visualizer, curve=None): + self.visualizer = visualizer + self.clear() + + self.nVertices = 0 + self.nEdges = 0 + self.vertexDegree = [] + self.vertices_old = {} + self.vertices = [] + self.edges_old = {} + self.nEdges = 0 + if curve is None: + self.networkCurve = NetworkCurve(self) + else: + self.networkCurve = curve + + self.edges = [] + self.minEdgeWeight = sys.maxint + self.maxEdgeWeight = 0 + + if visualizer == None: + xMin = self.axisScaleDiv(QwtPlot.xBottom).interval().minValue() + xMax = self.axisScaleDiv(QwtPlot.xBottom).interval().maxValue() + yMin = self.axisScaleDiv(QwtPlot.yLeft).interval().minValue() + yMax = self.axisScaleDiv(QwtPlot.yLeft).interval().maxValue() + self.addMarker("no network", (xMax - xMin) / 2, (yMax - yMin) / 2, alignment=Qt.AlignCenter, size=self.fontSize) + self.tooltipNeighbours = 0 + self.replot() + return + + self.nVertices = visualizer.graph.nVertices + #add nodes + for v in range(0, self.nVertices): + self.vertices_old[v] = (None, []) + vertex = NetworkVertex() + vertex.index = v + self.vertices.append(vertex) + + #build edge index + row_ind = {} + if visualizer.graph.links != None and len(visualizer.graph.links) > 0: + for i, r in enumerate(visualizer.graph.links): + u = int(r['u'].value) + v = int(r['v'].value) + u_dict = row_ind.get(u, {}) + v_dict = row_ind.get(v, {}) + u_dict[v] = i + v_dict[u] = i + row_ind[u] = u_dict + row_ind[v] = v_dict + + #add edges + for (i, j) in visualizer.graph.getEdges(): + self.edges_old[self.nEdges] = (None, i, j) + edge = NetworkEdge() + edge.u = self.vertices[i] + edge.v = self.vertices[j] + + edge.weight = float(str(visualizer.graph[i, j][0])) + + #print "weight:", edge.weight + + self.edges.append(edge) + self.nEdges += 1 + + if edge.weight != None and self.minEdgeWeight > edge.weight: + self.minEdgeWeight = edge.weight + + elif edge.weight != None and self.maxEdgeWeight < edge.weight: + self.maxEdgeWeight = edge.weight + + if visualizer.graph.directed: + edge.arrowu = 0 + edge.arrowv = 1 + + if not hasattr(visualizer.graph, 'links'): + visualizer.graph.links = None + + if visualizer.graph.links != None and len(visualizer.graph.links) > 0: + edge.links_index = row_ind[i + 1][j + 1] + row = visualizer.graph.links[edge.links_index] + edge.label = [str(row[r].value) for r in range(2, len(row))] + + #indexes = [k for k, x in enumerate(visualizer.graph.links) if (str(int(x[0])) == str(j + 1) and str(int(x[1])) == str(int(i + 1)))] + #for k in range(2, len(row[0])): + # edge.label.append(str(row[0][k])) + #else: + # print i, j, "not found" + + #if len(indexes) == 1: + # edge.links_index = indexes[0] + + if self.maxEdgeWeight < 10: + self.maxEdgeSize = self.maxEdgeWeight + else: + self.maxEdgeSize = 10 + + self.setEdgesSize() + self.setVerticesSize() + + self.networkCurve.coors = visualizer.network.coors + self.networkCurve.vertices = self.vertices + self.networkCurve.edges = self.edges + self.networkCurve.changed() + + def setEdgesSize(self): + if self.maxEdgeWeight > self.minEdgeWeight: + #print 'maxEdgeSize',self.maxEdgeSize + #print 'maxEdgeWeight',self.maxEdgeWeight + #print 'minEdgeWeight',self.minEdgeWeight + k = (self.maxEdgeSize - 1) / (self.maxEdgeWeight - self.minEdgeWeight) + for edge in self.edges: + if edge.weight == None: + size = 1 + edge.pen = QPen(edge.pen.color(), size) + edge.pen.setCapStyle(Qt.RoundCap) + else: + if self.invertEdgeSize: + size = (self.maxEdgeWeight - edge.weight - self.minEdgeWeight) * k + 1 + else: + size = (edge.weight - self.minEdgeWeight) * k + 1 + edge.pen = QPen(edge.pen.color(), size) + edge.pen.setCapStyle(Qt.RoundCap) + else: + for edge in self.edges: + edge.pen = QPen(edge.pen.color(), 1) + edge.pen.setCapStyle(Qt.RoundCap) + + def setVerticesSize(self, column=None, inverted=0): + if self.visualizer == None or self.visualizer.graph == None or self.visualizer.graph.items == None: + return + + column = str(column) + + if column in self.visualizer.graph.items.domain or (column.startswith("num of ") and column.replace("num of ", "") in self.visualizer.graph.items.domain): + values = [] + + if column in self.visualizer.graph.items.domain: + values = [x[column].value for x in self.visualizer.graph.items if not x[column].isSpecial()] + else: + values = [len(x[column.replace("num of ", "")].value.split(',')) for x in self.visualizer.graph.items] + + minVertexWeight = float(min(values or [0])) + maxVertexWeight = float(max(values or [0])) + + if maxVertexWeight - minVertexWeight == 0: + k = 1 #doesn't matter + else: + k = (self.maxVertexSize - self.minVertexSize) / (maxVertexWeight - minVertexWeight) + + def getValue(v): + if v.isSpecial(): + return minVertexWeight + else: + return float(v) + + if inverted: + for vertex in self.vertices: + if column in self.visualizer.graph.items.domain: + vertex.size = self.maxVertexSize - ((getValue(self.visualizer.graph.items[vertex.index][column]) - minVertexWeight) * k) + else: + vertex.size = self.maxVertexSize - ((len(self.visualizer.graph.items[vertex.index][column.replace("num of ", "")].value.split(',')) - minVertexWeight) * k) + + + vertex.pen.setWidthF(1 + float(vertex.size) / 20) + else: + for vertex in self.vertices: + if column in self.visualizer.graph.items.domain: + vertex.size = (getValue(self.visualizer.graph.items[vertex.index][column]) - minVertexWeight) * k + self.minVertexSize + else: + vertex.size = (float(len(self.visualizer.graph.items[vertex.index][column.replace("num of ", "")].value.split(','))) - minVertexWeight) * k + self.minVertexSize + + #print vertex.size + vertex.pen.setWidthF(1 + float(vertex.size) / 20) + else: + for vertex in self.vertices: + vertex.size = self.maxVertexSize + vertex.pen.setWidthF(1 + float(vertex.size) / 20) + + def updateCanvas(self): + self.setAxisAutoScale(self.xBottom) + self.setAxisAutoScale(self.yLeft) + self.updateData() + self.replot() + + def zoomExtent(self): + self.setAxisAutoScale(self.xBottom) + self.setAxisAutoScale(self.yLeft) + self.replot() + + def zoomSelection(self): + selection = self.networkCurve.getSelectedVertices() + if len(selection) > 0: + x = [self.visualizer.network.coors[0][v] for v in selection] + y = [self.visualizer.network.coors[1][v] for v in selection] + + oldXMin = self.axisScaleDiv(QwtPlot.xBottom).interval().minValue() + oldXMax = self.axisScaleDiv(QwtPlot.xBottom).interval().maxValue() + oldYMin = self.axisScaleDiv(QwtPlot.yLeft).interval().minValue() + oldYMax = self.axisScaleDiv(QwtPlot.yLeft).interval().maxValue() + newXMin = min(x) + newXMax = max(x) + newYMin = min(y) + newYMax = max(y) + self.zoomStack.append((oldXMin, oldXMax, oldYMin, oldYMax)) + self.setAxisScale(QwtPlot.xBottom, newXMin - 100, newXMax + 100) + self.setAxisScale(QwtPlot.yLeft, newYMin - 100, newYMax + 100) + self.replot() + Index: Orange/OrangeWidgets/Unsupervised/OWNetworkFile.py =================================================================== --- Orange/OrangeWidgets/Unsupervised/OWNetworkFile.py (revision 10466) +++ Orange/OrangeWidgets/Unsupervised/OWNetworkFile.py (revision 10466) @@ -0,0 +1,452 @@ +""" +THIS WIDGET IS OBSOLETE; USE OWNxFile.py +""" + +# +# OWNetworkFile.py +# The Network File Widget +# A widget for opening Network related files +# +import OWGUI, string, os.path, user, sys +import orngNetwork + +from OWWidget import * +from orange import Graph +from orange import GraphAsList + +class OWNetworkFile(OWWidget): + + settingsList=["recentFiles", "recentDataFiles", "recentEdgesFiles"] + + def __init__(self,parent=None, signalManager = None): + OWWidget.__init__(self, parent, signalManager, "Network File", wantMainArea=False) + + self.inputs = [] + self.outputs = [("Network", orngNetwork.Network), ("Items", ExampleTable)] + + #set default settings + self.recentFiles = ["(none)"] + self.recentDataFiles = ["(none)"] + self.recentEdgesFiles = ["(none)"] + + self.domain = None + self.graph = None + #get settings from the ini file, if they exist + self.loadSettings() + + #GUI + self.controlArea.layout().setMargin(4) + self.box = OWGUI.widgetBox(self.controlArea, box = "Graph File", orientation = "horizontal") + self.filecombo = OWGUI.comboBox(self.box, self, "filename") + self.filecombo.setMinimumWidth(250) + button = OWGUI.button(self.box, self, '...', callback = self.browseNetFile, disabled=0) + button.setIcon(self.style().standardIcon(QStyle.SP_DirOpenIcon)) + button.setSizePolicy(QSizePolicy.Maximum, QSizePolicy.Fixed) + + self.databox = OWGUI.widgetBox(self.controlArea, box = "Vertices Data File", orientation = "horizontal") + self.datacombo = OWGUI.comboBox(self.databox, self, "dataname") + self.datacombo.setMinimumWidth(250) + button = OWGUI.button(self.databox, self, '...', callback = self.browseDataFile, disabled=0) + button.setIcon(self.style().standardIcon(QStyle.SP_DirOpenIcon)) + button.setSizePolicy(QSizePolicy.Maximum, QSizePolicy.Fixed) + + self.edgesbox = OWGUI.widgetBox(self.controlArea, box = "Edges Data File", orientation = "horizontal") + self.edgescombo = OWGUI.comboBox(self.edgesbox, self, "edgesname") + self.edgescombo.setMinimumWidth(250) + button = OWGUI.button(self.edgesbox, self, '...', callback = self.browseEdgesFile, disabled=0) + button.setIcon(self.style().standardIcon(QStyle.SP_DirOpenIcon)) + button.setSizePolicy(QSizePolicy.Maximum, QSizePolicy.Fixed) + + # info + box = OWGUI.widgetBox(self.controlArea, "Info") + self.infoa = OWGUI.widgetLabel(box, 'No data loaded.') + self.infob = OWGUI.widgetLabel(box, ' ') + self.infoc = OWGUI.widgetLabel(box, ' ') + self.infod = OWGUI.widgetLabel(box, ' ') + + OWGUI.rubber(self.controlArea) + self.resize(150,100) + self.activateLoadedSettings() + + # connecting GUI to code + self.connect(self.filecombo, SIGNAL('activated(int)'), self.selectNetFile) + self.connect(self.datacombo, SIGNAL('activated(int)'), self.selectDataFile) + self.connect(self.edgescombo, SIGNAL('activated(int)'), self.selectEdgesFile) + + # set the comboboxes + def setFileLists(self): + self.filecombo.clear() + if not self.recentFiles: + self.filecombo.addItem("(none)") + for file in self.recentFiles: + if file == "(none)": + self.filecombo.addItem("(none)") + else: + self.filecombo.addItem(os.path.split(file)[1]) + self.filecombo.addItem("Browse documentation networks...") + + self.datacombo.clear() + if not self.recentDataFiles: + self.datacombo.addItem("(none)") + for file in self.recentDataFiles: + if file == "(none)": + self.datacombo.addItem("(none)") + else: + self.datacombo.addItem(os.path.split(file)[1]) + + self.edgescombo.clear() + if not self.recentEdgesFiles: + self.edgescombo.addItem("(none)") + for file in self.recentEdgesFiles: + if file == "(none)": + self.edgescombo.addItem("(none)") + else: + self.edgescombo.addItem(os.path.split(file)[1]) + + self.filecombo.updateGeometry() + self.datacombo.updateGeometry() + self.edgescombo.updateGeometry() + + def activateLoadedSettings(self): + # remove missing data set names + self.recentFiles = filter(os.path.exists, self.recentFiles) + self.recentDataFiles = filter(os.path.exists, self.recentDataFiles) + self.recentEdgesFiles = filter(os.path.exists, self.recentEdgesFiles) + + self.recentFiles.append("(none)") + self.recentDataFiles.append("(none)") + self.recentEdgesFiles.append("(none)") + self.setFileLists() + + if len(self.recentFiles) > 0 and os.path.exists(self.recentFiles[0]): + self.selectNetFile(0) + + if len(self.recentDataFiles) > 1 and os.path.exists(self.recentDataFiles[1]): + self.selectDataFile(1) + + if len(self.recentEdgesFiles) > 1 and os.path.exists(self.recentEdgesFiles[1]): + self.selectEdgesFile(1) + + # user selected a graph file from the combo box + def selectNetFile(self, n): + if n < len(self.recentFiles) : + name = self.recentFiles[n] + self.recentFiles.remove(name) + self.recentFiles.insert(0, name) + elif n: + self.browseNetFile(1) + + if len(self.recentFiles) > 0: + self.setFileLists() + self.openFile(self.recentFiles[0]) + + # user selected a data file from the combo box + def selectEdgesFile(self, n): + if n < len(self.recentEdgesFiles) : + name = self.recentEdgesFiles[n] + self.recentEdgesFiles.remove(name) + self.recentEdgesFiles.insert(0, name) + + if len(self.recentEdgesFiles) > 0: + self.setFileLists() + self.addEdgesFile(self.recentEdgesFiles[0]) + + def selectDataFile(self, n): + if n < len(self.recentDataFiles) : + name = self.recentDataFiles[n] + self.recentDataFiles.remove(name) + self.recentDataFiles.insert(0, name) + + if len(self.recentDataFiles) > 0: + self.setFileLists() + self.addDataFile(self.recentDataFiles[0]) + + def openFile(self, fn): + """Read network from file.""" + + # read network file + if fn != "(none)": + fileExt = lower(os.path.splitext(fn)[1]) + if not fileExt in (".net", ".gml"): + self.graph = None + self.send("Network", None) + self.send("Items", None) + self.infoa.setText('No data loaded') + self.infob.setText('File extension does not match') + self.infoc.setText('') + self.infod.setText('') + return + + data = self.readNetFile(fn) + + if data == None: + self.graph = None + self.send("Network", None) + self.send("Items", None) + self.infoa.setText('No data loaded') + self.infob.setText('Error reading file') + self.infoc.setText('') + self.infod.setText('') + return + + self.infoa.setText("%d nodes" % data.nVertices) + + if data.directed: + self.infob.setText("Directed graph") + else: + self.infob.setText("Undirected graph") + + # make new data and send it + fName = os.path.split(fn)[1] + if "." in fName: + #data.name = string.join(string.split(fName, '.')[:-1], '.') + pass + else: + #data.name = fName + pass + + self.graph = data + else: + self.graph = None + self.send("Network", None) + self.send("Items", None) + self.infoa.setText('No data loaded') + self.infob.setText('') + self.infoc.setText('') + self.infod.setText('') + + # Find items data file for selected network + items_candidate = os.path.splitext(fn)[0] + ".tab" + if os.path.exists(items_candidate): + self.readDataFile(items_candidate) + self.recentDataFiles.insert(0, items_candidate) + elif os.path.exists(os.path.splitext(fn)[0] + "_items.tab"): + items_candidate = os.path.splitext(fn)[0] + "_items.tab" + self.readDataFile(items_candidate) + self.recentDataFiles.insert(0, items_candidate) + else: + if '(none)' in self.recentDataFiles: + self.recentDataFiles.remove('(none)') + self.recentDataFiles.insert(0, '(none)') + + # Find links data file for selected network + links_candidate = os.path.splitext(fn)[0] + "_links.tab" + if os.path.exists(links_candidate): + self.readEdgesFile(links_candidate) + self.recentEdgesFiles.insert(0, links_candidate) + else: + if '(none)' in self.recentEdgesFiles: + self.recentEdgesFiles.remove('(none)') + self.recentEdgesFiles.insert(0, '(none)') + + self.setFileLists() + + self.send("Network", self.graph) + if self.graph != None and self.graph.items != None: + self.send("Items", self.graph.items) + else: + self.send("Items", None) + + def addDataFile(self, fn): + if fn == "(none)" or self.graph == None: + self.infoc.setText("No vertices data file specified") + self.send("Network", None) + self.send("Items", None) + return + + self.readDataFile(fn) + + self.send("Network", self.graph) + self.send("Items", self.graph.items) + + def readDataFile(self, fn): + table = ExampleTable(fn) + + if len(table) != self.graph.nVertices: + self.infoc.setText("Vertices data length does not match number of vertices") + + if '(none)' in self.recentDataFiles: + self.recentDataFiles.remove('(none)') + + self.recentDataFiles.insert(0, '(none)') + self.setFileLists() + return + + self.graph.setattr("items", table) + self.infoc.setText("Vertices data file added") + + def addEdgesFile(self, fn): + if fn == "(none)" or self.graph == None: + self.infod.setText("No edges data file specified") + #self.graph.setattr("links", None) + self.send("Network", self.graph) + self.send("Items", None) + return + + self.readEdgesFile(fn) + + self.send("Network", self.graph) + self.send("Items", self.graph.items) + + def readEdgesFile(self, fn): + table = ExampleTable(fn) + if self.graph.directed: + nEdges = len(self.graph.getEdges()) + else: + nEdges = len(self.graph.getEdges()) / 2 + + if len(table) != nEdges: + self.infod.setText("Edges data length does not match number of edges") + + if '(none)' in self.recentEdgesFiles: + self.recentEdgesFiles.remove('(none)') + + self.recentEdgesFiles.insert(0, '(none)') + self.setFileLists() + return + + self.graph.setattr("links", table) + self.infod.setText("Edges data file added") + + def browseNetFile(self, inDemos=0): + """user pressed the '...' button to manually select a file to load""" + + "Display a FileDialog and select a file" + if inDemos: + import os + try: + import orngConfiguration + startfile = orngConfiguration.datasetsPath + except: + startfile = "" + + if not startfile or not os.path.exists(startfile): + try: + import win32api, win32con + t = win32api.RegOpenKey(win32con.HKEY_LOCAL_MACHINE, "SOFTWARE\\Python\\PythonCore\\%i.%i\\PythonPath\\Orange" % sys.version_info[:2], 0, win32con.KEY_READ) + t = win32api.RegQueryValueEx(t, "")[0] + startfile = t[:t.find("orange")] + "orange\\doc\\networks" + except: + startfile = "" + + if not startfile or not os.path.exists(startfile): + d = OWGUI.__file__ + if d[-8:] == "OWGUI.py": + startfile = d[:-22] + "doc/networks" + elif d[-9:] == "OWGUI.pyc": + startfile = d[:-23] + "doc/networks" + + if not startfile or not os.path.exists(startfile): + d = os.getcwd() + if d[-12:] == "OrangeCanvas": + startfile = d[:-12]+"doc/networks" + else: + if d[-1] not in ["/", "\\"]: + d+= "/" + startfile = d+"doc/networks" + + if not os.path.exists(startfile): + QMessageBox.information( None, "File", "Cannot find the directory with example networks", QMessageBox.Ok + QMessageBox.Default) + return + else: + if len(self.recentFiles) == 0 or self.recentFiles[0] == "(none)": + if sys.platform == "darwin": + startfile = user.home + else: + startfile = "." + else: + startfile = self.recentFiles[0] + + filename = unicode(QFileDialog.getOpenFileName(self, 'Open a Network File', + startfile, "Pajek files (*.net)\nGML files (*.gml)\nAll files (*.*)")) + + if filename == "": return + if filename in self.recentFiles: self.recentFiles.remove(filename) + self.recentFiles.insert(0, filename) + self.setFileLists() + self.selectNetFile(0) + + def browseDataFile(self, inDemos=0): + if self.graph == None: + return + + #Display a FileDialog and select a file + if len(self.recentDataFiles) == 0 or self.recentDataFiles[0] == "(none)": + if len(self.recentFiles) == 0 or self.recentFiles[0] == "(none)": + if sys.platform == "darwin": + startfile = user.home + else: + startfile="." + else: + startfile = os.path.dirname(self.recentFiles[0]) + + else: + startfile = self.recentDataFiles[0] + + filename = unicode(QFileDialog.getOpenFileName(self, 'Open a Vertices Data File', startfile, 'Data files (*.tab)\nAll files(*.*)')) + + if filename == "": return + if filename in self.recentDataFiles: self.recentDataFiles.remove(filename) + self.recentDataFiles.insert(0, filename) + self.setFileLists() + self.addDataFile(self.recentDataFiles[0]) + + def browseEdgesFile(self, inDemos=0): + if self.graph == None: + return + + #Display a FileDialog and select a file + if len(self.recentEdgesFiles) == 0 or self.recentEdgesFiles[0] == "(none)": + if len(self.recentFiles) == 0 or self.recentFiles[0] == "(none)": + if sys.platform == "darwin": + startfile = user.home + else: + startfile="." + else: + startfile = os.path.dirname(self.recentFiles[0]) + + else: + startfile = self.recentEdgesFiles[0] + + filename = unicode(QFileDialog.getOpenFileName(self, 'Open a Edges Data File', startfile, 'Data files (*.tab)\nAll files(*.*)')) + + if filename == "": return + if filename in self.recentEdgesFiles: self.recentEdgesFiles.remove(filename) + self.recentEdgesFiles.insert(0, filename) + self.setFileLists() + self.addEdgesFile(self.recentEdgesFiles[0]) + + def setInfo(self, info): + for (i, s) in enumerate(info): + self.info[i].setText(s) + + def readNetFile(self, fn): + try: + net = orngNetwork.Network.read(fn) + self.infoc.setText("Vertices data generated and added automatically") + except: + self.infoa.setText("Could not read file") + self.infob.setText("") + self.infoc.setText("") + self.infod.setText("") + return None + + return net + + def sendReport(self): + self.reportSettings("Network file", + [("File name", self.filecombo.currentText()), + ("Vertices", self.graph.nVertices), + hasattr(self.graph, "directed") and ("Directed", OWGUI.YesNo[self.graph.directed])]) + self.reportSettings("Vertices meta data", [("File name", self.datacombo.currentText())]) + self.reportData(self.graph.items, None) + self.reportSettings("Edges meta data", [("File name", self.edgescombo.currentText())]) + self.reportData(self.graph.links, None, None) + +if __name__ == "__main__": + a=QApplication(sys.argv) + owf=OWNetworkFile() + owf.activateLoadedSettings() + owf.show() + a.exec_() + owf.saveSettings() Index: Orange/OrangeWidgets/Unsupervised/OWNetworkFromDistances.py =================================================================== --- Orange/OrangeWidgets/Unsupervised/OWNetworkFromDistances.py (revision 9671) +++ Orange/OrangeWidgets/Unsupervised/OWNetworkFromDistances.py (revision 9671) @@ -0,0 +1,85 @@ +""" +THIS WIDGET IS OBSOLETE; USE OWNxFromDistances.py +""" + +# +# OWNetworkFromDistances.py +# + +import OWGUI +import orange +import orngNetwork +import copy, random + +from OWNetworkHist import * +from OWWidget import * +from OWGraph import * +from OWHist import * + +class OWNetworkFromDistances(OWWidget, OWNetworkHist): + settingsList=["spinLowerThreshold", "spinUpperThreshold", "netOption", "dstWeight", "kNN", "andor", "excludeLimit"] + + def __init__(self, parent=None, signalManager=None): + OWWidget.__init__(self, parent, signalManager, "Network from Distances") + OWNetworkHist.__init__(self) + + self.inputs = [("Distances", orange.SymMatrix, self.setMatrix)] + self.outputs = [("Network", orngNetwork.Network), ("Data", ExampleTable), ("Distances", orange.SymMatrix)] + + self.addHistogramControls() + + # get settings from the ini file, if they exist + self.loadSettings() + + # GUI + # general settings + boxHistogram = OWGUI.widgetBox(self.mainArea, box = "Distance histogram") + self.histogram = OWHist(self, boxHistogram) + boxHistogram.layout().addWidget(self.histogram) + + boxHistogram.setMinimumWidth(500) + boxHistogram.setMinimumHeight(300) + + # info + boxInfo = OWGUI.widgetBox(self.controlArea, box = "Network info") + self.infoa = OWGUI.widgetLabel(boxInfo, "No data loaded.") + self.infob = OWGUI.widgetLabel(boxInfo, '') + self.infoc = OWGUI.widgetLabel(boxInfo, '') + + OWGUI.rubber(self.controlArea) + + self.resize(700, 100) + + def sendReport(self): + self.reportSettings("Settings", + [("Edge thresholds", "%.5f - %.5f" % (self.spinLowerThreshold, self.spinUpperThreshold)), + ("Selected vertices", ["All", "Without isolated vertices", "Largest component", "Connected with vertex"][self.netOption]), + ("Weight", ["Distance", "1 - Distance"][self.dstWeight])]) + self.reportSection("Histogram") + self.reportImage(self.histogram.saveToFileDirect, QSize(400,300)) + self.reportSettings("Output graph", + [("Vertices", self.matrix.dim), + ("Edges", self.nedges), + ("Connected vertices", "%i (%.1f%%)" % (self.pconnected, self.pconnected / max(1, float(self.matrix.dim))*100)), + ]) + + def sendSignals(self): + if self.graph != None: + #setattr(matrix, "items", self.graph.items) + self.matrix.items = self.graph.items + + self.send("Network", self.graph) + + if self.matrix: + self.send("Distances", self.matrix) + + if self.graph == None: + self.send("Data", None) + else: + self.send("Data", self.graph.items) + +if __name__ == "__main__": + appl = QApplication(sys.argv) + ow = OWNetworkFromDistances() + ow.show() + appl.exec_() Index: Orange/OrangeWidgets/Unsupervised/OWNxAnalysis.py =================================================================== --- Orange/OrangeWidgets/Unsupervised/OWNxAnalysis.py (revision 10510) +++ Orange/OrangeWidgets/Unsupervised/OWNxAnalysis.py (revision 10046) @@ -6,5 +6,4 @@ 6425 """ -from PyQt4.QtCore import QMutex import numpy import networkx as nx @@ -67,5 +66,5 @@ self.outputs = [("Network", Orange.network.Graph), ("Items", Orange.data.Table)] - + self.methods = [ ("number_of_nodes", True, "Number of nodes", GRAPHLEVEL, lambda G: G.number_of_nodes()), @@ -76,17 +75,10 @@ ("average_shortest_path_length", False, "Average shortest path length", GRAPHLEVEL, nx.average_shortest_path_length), ("density", True, "Density", GRAPHLEVEL, nx.density), - ("degree_assortativity_coefficient", False, \ - "Degree assortativity coefficient", GRAPHLEVEL, \ - nx.degree_assortativity_coefficient if \ - hasattr(nx, "degree_assortativity_coefficient") else None), + ("degree_assortativity_coefficient", False, "Degree assortativity coefficient", GRAPHLEVEL, nx.degree_assortativity_coefficient), # additional attr needed #("attribute_assortativity_coefficient", False, "Attribute assortativity coefficient", GRAPHLEVEL, nx.attribute_assortativity_coefficient), #("numeric_assortativity_coefficient", False, "Numeric assortativity coefficient", GRAPHLEVEL, nx.numeric_assortativity_coefficient), - ("degree_pearson_correlation_coefficient", False, \ - "Degree pearson correlation coefficient", GRAPHLEVEL, \ - nx.degree_pearson_correlation_coefficient if\ - hasattr(nx, "degree_pearson_correlation_coefficient") else None), - ("estrada_index", False, "Estrada index", GRAPHLEVEL, \ - nx.estrada_index if hasattr(nx, "estrada_index") else None), + ("degree_pearson_correlation_coefficient", False, "Degree pearson correlation coefficient", GRAPHLEVEL, nx.degree_pearson_correlation_coefficient), + ("estrada_index", False, "Estrada index", GRAPHLEVEL, nx.estrada_index), ("graph_clique_number", False, "Graph clique number", GRAPHLEVEL, nx.graph_clique_number), ("graph_number_of_cliques", False, "Graph number of cliques", GRAPHLEVEL, nx.graph_number_of_cliques), @@ -120,9 +112,5 @@ ("current_flow_closeness_centrality", False, "Information centrality", NODELEVEL, nx.current_flow_closeness_centrality), ("current_flow_betweenness_centrality", False, "Random-walk betweenness centrality", NODELEVEL, nx.current_flow_betweenness_centrality), - ("approximate_current_flow_betweenness_centrality", False, \ - "Approx. random-walk betweenness centrality", NODELEVEL, \ - nx.approximate_current_flow_betweenness_centrality if \ - hasattr(nx, "approximate_current_flow_betweenness_centrality") \ - else None), + ("approximate_current_flow_betweenness_centrality", False, "Approx. random-walk betweenness centrality", NODELEVEL, nx.approximate_current_flow_betweenness_centrality), ("eigenvector_centrality", False, "Eigenvector centrality", NODELEVEL, nx.eigenvector_centrality), ("eigenvector_centrality_numpy", False, "Eigenvector centrality (NumPy)", NODELEVEL, nx.eigenvector_centrality_numpy), @@ -132,6 +120,4 @@ ("closeness_vitality", False, "Closeness vitality", NODELEVEL, nx.closeness_vitality), ] - - self.methods = [method for method in self.methods if method[-1] is not None] self.auto_commit = False Index: Orange/__init__.py =================================================================== --- Orange/__init__.py (revision 10540) +++ Orange/__init__.py (revision 10491) @@ -80,5 +80,4 @@ _import("ensemble.boosting") _import("ensemble.forest") -_import("ensemble.stacking") _import("regression") Index: Orange/classification/rules.py =================================================================== --- Orange/classification/rules.py (revision 10519) +++ Orange/classification/rules.py (revision 10407) @@ -39,11 +39,36 @@ Validator_LRS = Orange.core.RuleValidator_LRS -from Orange.orng.orngABML import \ - ArgumentFilter_hasSpecial, \ - create_dichotomous_class, \ - evaluateAndSortArguments from Orange.misc import deprecated_keywords from Orange.misc import deprecated_members + +class ConvertClass: + """ Converting class variables into dichotomous class variable. """ + def __init__(self, classAtt, classValue, newClassAtt): + self.classAtt = classAtt + self.classValue = classValue + self.newClassAtt = newClassAtt + + def __call__(self, example, returnWhat): + if example[self.classAtt] == self.classValue: + return Orange.data.Value(self.newClassAtt, self.classValue + "_") + else: + return Orange.data.Value(self.newClassAtt, "not " + self.classValue) + + +def create_dichotomous_class(domain, att, value, negate, removeAtt=None): + # create new variable + newClass = Orange.feature.Discrete(att.name + "_", values=[str(value) + "_", "not " + str(value)]) + positive = Orange.data.Value(newClass, str(value) + "_") + negative = Orange.data.Value(newClass, "not " + str(value)) + newClass.getValueFrom = ConvertClass(att, str(value), newClass) + + att = [a for a in domain.attributes] + newDomain = Orange.data.Domain(att + [newClass]) + newDomain.addmetas(domain.getmetas()) + if negate == 1: + return (newDomain, negative) + else: + return (newDomain, positive) @@ -534,5 +559,5 @@ rule_sig=1.0, att_sig=1.0, postpruning=None, min_quality=0., min_coverage=1, min_improved=1, min_improved_perc=0.0, learn_for_class=None, learn_one_rule=False, evd=None, evd_arguments=None, prune_arguments=False, analyse_argument= -1, - alternative_learner=None, min_cl_sig=0.5, min_beta=0.0, set_prefix_rules=False, add_sub_rules=True, debug=False, + alternative_learner=None, min_cl_sig=0.5, min_beta=0.0, set_prefix_rules=False, add_sub_rules=False, debug=False, **kwds): @@ -627,5 +652,5 @@ while aes: if self.analyse_argument > -1 and \ - (isinstance(self.analyse_argument, Orange.core.Example) and not Orange.core.Example(dich_data.domain, self.analyse_argument) == aes[0] or \ + (isinstance(self.analyse_argument, Orange.data.Instance) and not Orange.data.Instance(dich_data.domain, self.analyse_argument) == aes[0] or \ isinstance(self.analyse_argument, int) and not dich_data[self.analyse_argument] == aes[0]): aes = aes[1:] @@ -642,4 +667,5 @@ else: aes = aes[1:] + aes = aes[1:] if not progress and self.debug: @@ -648,5 +674,5 @@ # remove all examples covered by rules for rule in rules: - dich_data = self.remove_covered_examples(rule, dich_data, weight_id, True) + dich_data = self.remove_covered_examples(rule, dich_data, weight_id) if progress: progress(self.remaining_probability(dich_data), None) @@ -657,19 +683,13 @@ while dich_data: # learn a rule - rule, good_rule = self.learn_normal_rule(dich_data, weight_id, self.apriori) + rule = self.learn_normal_rule(dich_data, weight_id, self.apriori) if not rule: break if self.debug: - if good_rule: - print "rule learned: ", rule_to_string(rule), rule.quality - else: - print "rule only to influence learning: ", rule_to_string(rule), rule.quality - - dich_data = self.remove_covered_examples(rule, dich_data, weight_id, good_rule) - + print "rule learned: ", Orange.classification.rules.rule_to_string(rule), rule.quality + dich_data = self.remove_covered_examples(rule, dich_data, weight_id) if progress: progress(self.remaining_probability(dich_data), None) - if good_rule: - rules.append(rule) + rules.append(rule) if self.learn_one_rule: break @@ -695,7 +715,4 @@ if not positive_args: # something wrong raise "There is a problem with argumented example %s" % str(ae) - return None - if False in [p(ae) for p in positive_args]: # a positive argument is not covering this example - raise "One argument does not cover critical example: %s!"%str(ae) return None negative_args = self.init_neg_args(ae, examples, weight_id) @@ -790,20 +807,14 @@ if hasattr(self.rule_finder.evaluator, "bestRule"): self.rule_finder.evaluator.bestRule = None - rule = self.rule_finder(examples,weight_id,0,RuleList()) + rule = self.rule_finder(examples, weight_id, 0, RuleList()) if hasattr(self.rule_finder.evaluator, "bestRule") and self.rule_finder.evaluator.returnExpectedProb: - if not self.rule_finder.evaluator.bestRule and rule.quality > 0: - return (rule, False) rule = self.rule_finder.evaluator.bestRule self.rule_finder.evaluator.bestRule = None if self.postpruning: - rule = self.postpruning(rule,examples,weight_id,0, aprior) - return (rule, True) - - - def remove_covered_examples(self, rule, examples, weight_id, good_rule): - if good_rule: - nexamples, nweight = self.cover_and_remove(rule, examples, weight_id, 0) - else: - nexamples, nweight = self.cover_and_remove.mark_examples_solved(rule,examples,weight_id,0) + rule = self.postpruning(rule, examples, weight_id, 0, aprior) + return rule + + def remove_covered_examples(self, rule, examples, weight_id): + nexamples, nweight = self.cover_and_remove(rule, examples, weight_id, 0) return nexamples @@ -848,5 +859,5 @@ self.rule_finder.evaluator.returnExpectedProb = oldREP tmpList.sort(lambda x, y:-cmp(x.quality, y.quality)) - tmpList = tmpList[:self.ruleFilter.width] + tmpList = tmpList[:self.rule_filter.width] for tmpRule in tmpList: @@ -862,5 +873,5 @@ tmpRule2.filterAndStore(examples, weight_id, r.classifier.default_val) tmpRule2.complexity += 1 - if tmpRule2.class_distribution.abs < tmpRule.class_distribution.abs: + if tmpRule2.class_distribution.abs < tmprule.class_distribution.abs: tmpList2.append(tmpRule2) tmpList = tmpList2 @@ -1214,7 +1225,7 @@ if r and not rule_in_set(r, best_rules) and int(examples[r_i].getclass()) == int(r.classifier.default_value): if hasattr(r.learner, "arg_example"): - r.setattr("best_example", r.learner.arg_example) + setattr(r, "best_example", r.learner.arg_example) else: - r.setattr("best_example", examples[r_i]) + setattr(r, "best_example", examples[r_i]) best_rules.append(r) return best_rules @@ -1223,21 +1234,44 @@ """ if example has an argument, then the rule must be consistent with the argument. """ example = getattr(rule.learner, "arg_example", None) - if example: - for ei, e in enumerate(examples): - if e == example: - e[self.prob_attribute] = rule.quality+0.001 # 0.001 is added to avoid numerical errors - self.best_rule[ei]=rule - else: - for ei, e in enumerate(examples): - if rule(e) and rule.quality>e[self.prob_attribute]: - e[self.prob_attribute] = rule.quality+0.001 # 0.001 is added to avoid numerical errors - self.best_rule[ei]=rule + for ei, e in enumerate(examples): + if e == example: + e[self.prob_attribute] = 1.0 + self.best_rule[ei] = rule + elif rule(e) and rule.quality > e[self.prob_attribute]: + e[self.prob_attribute] = rule.quality + 0.001 # 0.001 is added to avoid numerical errors + self.best_rule[ei] = rule return (examples, weights) - def mark_examples_solved(self, rule, examples, weights, target_class): - for ei, e in enumerate(examples): - if rule(e): - e[self.prob_attribute] = 1.0 - return (examples, weights) + def filter_covers_example(self, example, filter): + filter_indices = CoversArguments.filterIndices(filter) + if filter(example): + try: + if example[self.argument_id].value and len(example[self.argument_id].value.positive_arguments) > 0: # example has positive arguments + # conditions should cover at least one of the positive arguments + one_arg_covered = False + for pA in example[self.argument_id].value.positive_arguments: + arg_covered = [self.condIn(c, filter_indices) for c in pA.filter.conditions] + one_arg_covered = one_arg_covered or len(arg_covered) == sum(arg_covered) #arg_covered + if one_arg_covered: + break + if not one_arg_covered: + return False + if example[self.argument_id].value and len(example[self.argument_id].value.negative_arguments) > 0: # example has negative arguments + # condition should not cover neither of negative arguments + for pN in example[self.argument_id].value.negative_arguments: + arg_covered = [self.condIn(c, filter_indices) for c in pN.filter.conditions] + if len(arg_covered) == sum(arg_covered): + return False + except: + return True + return True + return False + + def condIn(self, cond, filter_indices): # is condition in the filter? + condInd = CoversArguments.conditionIndex(cond) + if operator.or_(condInd, filter_indices[cond.position]) == filter_indices[cond.position]: + return True + return False + def covered_percentage(self, examples): @@ -1708,52 +1742,58 @@ # This filter is the ugliest code ever! Problem is with Orange, I had some problems with inheriting deepCopy # I should take another look at it. -class ArgFilter(Orange.core.Filter): +class ArgFilter(Orange.data.filter.Filter): """ This class implements AB-covering principle. """ - def __init__(self, argument_id=None, filter = Orange.core.Filter_values(), arg_example = None): + def __init__(self, argument_id=None, filter=Orange.data.filter.Values(), arg_example=None): self.filter = filter - self.indices = getattr(filter,"indices",[]) - if not self.indices and len(filter.conditions)>0: - self.indices = RuleCoversArguments.filterIndices(filter) + self.indices = getattr(filter, "indices", []) + if not self.indices and len(filter.conditions) > 0: + self.indices = CoversArguments.filterIndices(filter) self.argument_id = argument_id self.domain = self.filter.domain self.conditions = filter.conditions self.arg_example = arg_example - self.only_arg_example = True - - def condIn(self,cond): # is condition in the filter? - condInd = RuleCoversArguments.conditionIndex(cond) - if operator.or_(condInd,self.indices[cond.position]) == self.indices[cond.position]: + + def condIn(self, cond): # is condition in the filter? + condInd = ruleCoversArguments.conditionIndex(cond) + if operator.or_(condInd, self.indices[cond.position]) == self.indices[cond.position]: return True return False - - def __call__(self,example): - if not self.filter(example): + + def __call__(self, example): +## print "in", self.filter(example)#, self.filter.conditions[0](example) +## print self.filter.conditions[1].values + if self.filter(example) and example != self.arg_example: + return True + elif self.filter(example): + try: + if example[self.argument_id].value and len(example[self.argument_id].value.positiveArguments) > 0: # example has positive arguments + # conditions should cover at least one of the positive arguments + oneArgCovered = False + for pA in example[self.argument_id].value.positiveArguments: + argCovered = [self.condIn(c) for c in pA.filter.conditions] + oneArgCovered = oneArgCovered or len(argCovered) == sum(argCovered) #argCovered + if oneArgCovered: + break + if not oneArgCovered: + return False + if example[self.argument_id].value and len(example[self.argument_id].value.negativeArguments) > 0: # example has negative arguments + # condition should not cover neither of negative arguments + for pN in example[self.argument_id].value.negativeArguments: + argCovered = [self.condIn(c) for c in pN.filter.conditions] + if len(argCovered) == sum(argCovered): + return False + except: + return True + return True + else: return False - elif (not self.only_arg_example or example == self.arg_example): - if example[self.argument_id].value and len(example[self.argument_id].value.positive_arguments)>0: # example has positive arguments - # conditions should cover at least one of the positive arguments - oneArgCovered = False - for pA in example[self.argument_id].value.positive_arguments: - argCovered = [self.condIn(c) for c in pA.filter.conditions] - oneArgCovered = oneArgCovered or len(argCovered) == sum(argCovered) #argCovered - if oneArgCovered: - break - if not oneArgCovered: - return False - if example[self.argument_id].value and len(example[self.argument_id].value.negative_arguments)>0: # example has negative arguments - # condition should not cover neither of negative arguments - for pN in example[self.argument_id].value.negative_arguments: - argCovered = [self.condIn(c) for c in pN.filter.conditions] - if len(argCovered)==sum(argCovered): - return False - return True - - def __setattr__(self,name,obj): - self.__dict__[name]=obj - self.filter.setattr(name,obj) + + def __setattr__(self, name, obj): + self.__dict__[name] = obj + self.filter.setattr(name, obj) def deep_copy(self): newFilter = ArgFilter(argument_id=self.argument_id) - newFilter.filter = Orange.core.Filter_values() #self.filter.deepCopy() + newFilter.filter = Orange.data.filter.Values() #self.filter.deepCopy() newFilter.filter.conditions = self.filter.conditions[:] newFilter.domain = self.filter.domain @@ -1763,6 +1803,6 @@ newFilter.conditions = newFilter.filter.conditions newFilter.indices = self.indices[:] - newFilter.arg_example = self.arg_example return newFilter + ArgFilter = deprecated_members({"argumentID": "argument_id"})(ArgFilter) @@ -1841,5 +1881,5 @@ PILAR (Probabilistic improvement of learning algorithms with rules). """ - def __init__(self, alternative_learner=None, min_cl_sig=0.5, min_beta=0.0, penalty=0.01, set_prefix_rules=False, optimize_betas=True): + def __init__(self, alternative_learner=None, min_cl_sig=0.5, min_beta=0.0, set_prefix_rules=False, optimize_betas=True): self.alternative_learner = alternative_learner self.min_cl_sig = min_cl_sig @@ -1848,5 +1888,4 @@ self.optimize_betas = optimize_betas self.selected_evaluation = CrossValidation(folds=5) - self.penalty = penalty def __call__(self, rules, examples, weight=0): @@ -1858,11 +1897,14 @@ ## for e in examples: ## prob_dist.append(classifier(e,Orange.core.GetProbabilities)) - cl = Orange.core.RuleClassifier_logit(rules, self.min_cl_sig, self.min_beta, self.penalty, examples, weight, self.set_prefix_rules, self.optimize_betas, classifier, prob_dist) + cl = RuleClassifier_logit(rules, self.min_cl_sig, self.min_beta, examples, weight, self.set_prefix_rules, self.optimize_betas, classifier, prob_dist) else: - cl = Orange.core.RuleClassifier_logit(rules, self.min_cl_sig, self.min_beta, self.penalty, examples, weight, self.set_prefix_rules, self.optimize_betas) - + cl = RuleClassifier_logit(rules, self.min_cl_sig, self.min_beta, examples, weight, self.set_prefix_rules, self.optimize_betas) + +## print "result" for ri, r in enumerate(cl.rules): cl.rules[ri].setattr("beta", cl.ruleBetas[ri]) - cl.setattr("all_rules", cl.rules) +## if cl.ruleBetas[ri] > 0: +## print Orange.classification.rules.rule_to_string(r), r.quality, cl.ruleBetas[ri] + cl.all_rules = cl.rules cl.rules = self.sort_rules(cl.rules) cl.ruleBetas = [r.beta for r in cl.rules] Index: Orange/data/io.py =================================================================== --- Orange/data/io.py (revision 10528) +++ Orange/data/io.py (revision 10255) @@ -480,6 +480,6 @@ elif cell == "": return variable.Descriptor - elif len(split_escaped_str(cell, " ")) > 1: - return variable.Discrete, split_escaped_str(cell, " ") + elif len(cell.split(",")) > 1: + return variable.Discrete, cell.split(",") else: raise ValueError("Unknown variable type definition %r." % cell) @@ -581,7 +581,5 @@ return False -def load_csv(file, create_new_on=MakeStatus.Incompatible, - delimiter=None, quotechar=None, escapechar=None, - has_header=None, has_types=None, has_annotations=None, **kwargs): +def load_csv(file, create_new_on=MakeStatus.Incompatible, **kwargs): """ Load an Orange.data.Table from s csv file. """ @@ -591,72 +589,41 @@ sample = file.read(5 * 2 ** 20) # max 5MB sample TODO: What if this is not enough. Try with a bigger sample dialect = snifer.sniff(sample) - - if has_header is None: - has_header = snifer.has_header(sample) - + has_header = snifer.has_header(sample) file.seek(0) # Rewind - - def kwparams(**kwargs): - """Return not None kwargs. - """ - return dict([(k, v) for k, v in kwargs.items() if v is not None]) - - fmtparam = kwparams(delimiter=delimiter, - quotechar=quotechar, - escapechar=escapechar) - - reader = csv.reader(file, dialect=dialect, - **fmtparam) + reader = csv.reader(file, dialect=dialect) header = types = var_attrs = None - row = first_row = reader.next() - - if has_header: - header = row - # Eat this row and move to the next - row = reader.next() - - # Guess types row - if has_types is None: - has_types = has_header and is_var_types_row(row) - - if has_types: - types = var_types(row) - # Eat this row and move to the next - row = reader.next() - - # Guess variable annotations row - if has_annotations is None: - has_annotations = has_header and has_types and \ - is_var_attributes_row(row) - - if has_annotations: - labels_row = row - var_attrs = var_attributes(row) - # Eat this row and move to the next - row = reader.next() - - if not header: - # Create a default header - header = ["F_%i" % i for i in range(len(first_row))] - +# if not has_header: +# raise ValueError("No header in the data file.") + + header = reader.next() + + if header: + # Try to get variable definitions + types_row = reader.next() + if is_var_types_row(types_row): + types = var_types(types_row) + + if types: + # Try to get the variable attributes + # (third line in the standard orange tab format). + labels_row = reader.next() + if is_var_attributes_row(labels_row): + var_attrs = var_attributes(labels_row) + + # If definitions not present fill with blanks if not types: - # Create blank variable types types = [None] * len(header) - if not var_attrs: - # Create blank variable attributes var_attrs = [None] * len(header) # start from the beginning file.seek(0) - reader = csv.reader(file, dialect=dialect, **fmtparam) - - for defined in [has_header, has_types, has_annotations]: - if defined: - # skip definition rows if present in the file + reader = csv.reader(file, dialect=dialect) + for defined in [header, types, var_attrs]: + if any(defined): # skip definition rows if present in the file reader.next() - + variables = [] undefined_vars = [] Index: Orange/ensemble/__init__.py =================================================================== --- Orange/ensemble/__init__.py (revision 10540) +++ Orange/ensemble/__init__.py (revision 9994) @@ -1,2 +1,188 @@ -__all__ = ["bagging", "boosting", "forest", "stacking"] +""" + +.. index:: ensemble + +Module Orange.ensemble implements Breiman's bagging and Random Forest, +and Freund and Schapire's boosting algorithms. + + +******* +Bagging +******* + +.. index:: bagging +.. index:: + single: ensemble; ensemble + +.. autoclass:: Orange.ensemble.bagging.BaggedLearner + :members: + :show-inheritance: + +.. autoclass:: Orange.ensemble.bagging.BaggedClassifier + :members: + :show-inheritance: + +******** +Boosting +******** + +.. index:: boosting +.. index:: + single: ensemble; boosting + + +.. autoclass:: Orange.ensemble.boosting.BoostedLearner + :members: + :show-inheritance: + +.. autoclass:: Orange.ensemble.boosting.BoostedClassifier + :members: + :show-inheritance: + +Example +======= +Let us try boosting and bagging on Lymphography data set and use TreeLearner +with post-pruning as a base learner. For testing, we use 10-fold cross +validation and observe classification accuracy. + +:download:`ensemble.py ` + +.. literalinclude:: code/ensemble.py + :lines: 7- + +Running this script, we may get something like:: + + Classification Accuracy: + tree: 0.764 + boosted tree: 0.770 + bagged tree: 0.790 + + +************* +Random Forest +************* + +.. index:: random forest +.. index:: + single: ensemble; random forest + +.. autoclass:: Orange.ensemble.forest.RandomForestLearner + :members: + :show-inheritance: + +.. autoclass:: Orange.ensemble.forest.RandomForestClassifier + :members: + :show-inheritance: + + +Example +======== + +The following script assembles a random forest learner and compares it +to a tree learner on a liver disorder (bupa) and housing data sets. + +:download:`ensemble-forest.py ` + +.. literalinclude:: code/ensemble-forest.py + :lines: 7- + +Notice that our forest contains 50 trees. Learners are compared through +3-fold cross validation:: + + Classification: bupa.tab + Learner CA Brier AUC + tree 0.586 0.829 0.575 + forest 0.710 0.392 0.752 + Regression: housing.tab + Learner MSE RSE R2 + tree 23.708 0.281 0.719 + forest 11.988 0.142 0.858 + +Perhaps the sole purpose of the following example is to show how to +access the individual classifiers once they are assembled into the +forest, and to show how we can assemble a tree learner to be used in +random forests. In the following example the best feature for decision +nodes is selected among three randomly chosen features, and maxDepth +and minExamples are both set to 5. + +:download:`ensemble-forest2.py ` + +.. literalinclude:: code/ensemble-forest2.py + :lines: 7- + +Running the above code would report on sizes (number of nodes) of the tree +in a constructed random forest. + + +Score Feature +============= + +L. Breiman (2001) suggested the possibility of using random forests as a +non-myopic measure of feature importance. + +The assessment of feature relevance with random forests is based on the +idea that randomly changing the value of an important feature greatly +affects instance's classification, while changing the value of an +unimportant feature does not affect it much. Implemented algorithm +accumulates feature scores over given number of trees. Importance of +all features for a single tree are computed as: correctly classified +OOB instances minus correctly classified OOB instances when the feature is +randomly shuffled. The accumulated feature scores are divided by the +number of used trees and multiplied by 100 before they are returned. + +.. autoclass:: Orange.ensemble.forest.ScoreFeature + :members: + +Computation of feature importance with random forests is rather slow and +importances for all features need to be computes simultaneously. When it +is called to compute a quality of certain feature, it computes qualities +for all features in the dataset. When called again, it uses the stored +results if the domain is still the same and the data table has not +changed (this is done by checking the data table's version and is +not foolproof; it will not detect if you change values of existing instances, +but will notice adding and removing instances; see the page on +:class:`Orange.data.Table` for details). + +:download:`ensemble-forest-measure.py ` + +.. literalinclude:: code/ensemble-forest-measure.py + :lines: 7- + +Corresponding output:: + + DATA:iris.tab + + first: 3.91, second: 0.38 + + different random seed + first: 3.39, second: 0.46 + + All importances: + sepal length: 3.39 + sepal width: 0.46 + petal length: 30.15 + petal width: 31.98 + +References +----------- +* L Breiman. Bagging Predictors. `Technical report No. 421 \ + `_. University of \ + California, Berkeley, 1994. +* Y Freund, RE Schapire. `Experiments with a New Boosting Algorithm \ + `_. Machine \ + Learning: Proceedings of the Thirteenth International Conference (ICML'96), 1996. +* JR Quinlan. `Boosting, bagging, and C4.5 \ + `_ . In Proc. of 13th \ + National Conference on Artificial Intelligence (AAAI'96). pp. 725-730, 1996. +* L Brieman. `Random Forests \ + `_.\ + Machine Learning, 45, 5-32, 2001. +* M Robnik-Sikonja. `Improving Random Forests \ + `_. In \ + Proc. of European Conference on Machine Learning (ECML 2004),\ + pp. 359-370, 2004. +""" + +__all__ = ["bagging", "boosting", "forest"] __docformat__ = 'restructuredtext' +import Orange.core as orange Index: Orange/ensemble/forest.py =================================================================== --- Orange/ensemble/forest.py (revision 10540) +++ Orange/ensemble/forest.py (revision 9919) @@ -34,6 +34,12 @@ split is (on average) as specified.""" - __new__ = Orange.misc._orange__new__(Orange.core.Learner) - + def __new__(cls, instances = None, weight_id = 0, **argkw): + self = Orange.core.Learner.__new__(cls, **argkw) + if instances: + self.__init__(**argkw) + return self.__call__(instances, weight_id) + else: + return self + def __init__(self, base, rand): self.base = base @@ -48,15 +54,13 @@ self.base.skip_prob, self.base.random_generator = osp, orand return r - _RandomForestSimpleTreeLearner = Orange.misc.deprecated_members({"weightID":"weight_id", "examples":"instances"})(_RandomForestSimpleTreeLearner) -class RandomForestLearner(Orange.core.Learner): - """ - Trains an ensemble predictor consisting of trees trained - on bootstrap - samples of training data. To increase - randomness, the tree learner considers only a subset of - candidate features at each node. The algorithm closely follows - the original procedure (Brieman, 2001) both in implementation and parameter +class RandomForestLearner(orange.Learner): + """ + Just like in bagging, classifiers in random forests are trained from bootstrap + samples of training data. Here, the classifiers are trees. However, to increase + randomness, at each node of the tree the best feature is + chosen from a subset of features in the data. We closely follow the + original algorithm (Brieman, 2001) both in implementation and parameter defaults. @@ -65,6 +69,6 @@ :param attributes: number of randomly drawn features among - which to select the best one to split the data sets - in tree nodes. The default, None, means the square root of + which to select the best to split the nodes in tree + induction. The default, None, means the square root of the number of features in the training data. Ignored if :obj:`learner` is specified. @@ -90,9 +94,8 @@ :param callback: a function to be called after every iteration of - induction of classifier. The call includes a parameter - (from 0.0 to 1.0) that provides an estimate - of completion of the learning progress. - - :param name: learner name. + induction of classifier. This is called with parameter + (from 0.0 to 1.0) that gives estimates on learning progress. + + :param name: name of the learner. :type name: string @@ -102,5 +105,11 @@ """ - __new__ = Orange.misc._orange__new__(Orange.core.Learner) + def __new__(cls, instances=None, weight = 0, **kwds): + self = orange.Learner.__new__(cls, **kwds) + if instances: + self.__init__(**kwds) + return self.__call__(instances, weight) + else: + return self def __init__(self, trees=100, attributes=None,\ @@ -476,7 +485,13 @@ a new split constructor. """ - - __new__ = Orange.misc._orange__new__(Orange.core.Learner) - + @deprecated_keywords({"weightID":"weight_id", "examples":"instances"}) + def __new__(cls, instances = None, weight_id = 0, **argkw): + self = Orange.core.Learner.__new__(cls, **argkw) + if instances: + self.__init__(**argkw) + return self.__call__(instances, weight_id) + else: + return self + def __init__(self, base, rand): self.base = base Index: range/ensemble/stacking.py =================================================================== --- Orange/ensemble/stacking.py (revision 10540) +++ (revision ) @@ -1,67 +1,0 @@ -import Orange - -class StackedClassificationLearner(Orange.classification.Learner): - """Stacking by inference of meta classifier from class probability estimates - on cross-validation held-out data for level-0 classifiers developed on held-in data sets. - - :param learners: level-0 learners. - :type learners: list - - :param meta_learner: meta learner (default: :class:`~Orange.classification.bayes.NaiveLearner`). - :type meta_learner: :class:`~Orange.classification.Learner` - - :param folds: number of iterations (folds) of cross-validation to assemble class probability data for meta learner. - - :param name: learner name (default: stacking). - :type name: string - - :rtype: :class:`~Orange.ensemble.stacking.StackedClassificationLearner` or - :class:`~Orange.ensemble.stacking.StackedClassifier` - """ - def __new__(cls, learners, data=None, weight=0, **kwds): - if data is None: - self = Orange.classification.Learner.__new__(cls) - return self - else: - self = cls(learners, **kwds) - return self(data, weight) - - def __init__(self, learners, meta_learner=Orange.classification.bayes.NaiveLearner(), folds=10, name='stacking'): - self.learners = learners - self.meta_learner = meta_learner - self.name = name - self.folds = folds - - def __call__(self, data, weight=0): - res = Orange.evaluation.testing.cross_validation(self.learners, data, self.folds) - features = [Orange.feature.Continuous("%d" % i) for i in range(len(self.learners) * (len(data.domain.class_var.values) - 1))] - domain = Orange.data.Domain(features + [data.domain.class_var]) - p_data = Orange.data.Table(domain) - for r in res.results: - p_data.append([p for ps in r.probabilities for p in ps[:-1]] + [r.actual_class]) - meta_classifier = self.meta_learner(p_data) - - classifiers = [l(data, weight) for l in self.learners] - feature_domain = Orange.data.Domain(features) - return StackedClassifier(classifiers, meta_classifier, name=self.name) - -class StackedClassifier: - """ - A classifier for stacking. Uses a set of level-0 classifiers to induce class probabilities, which - are an input to a meta-classifier to predict class probability for a given data instance. - - :param classifiers: a list of level-0 classifiers. - :type classifiers: list - - :param meta_classifier: meta-classifier. - :type meta_classifier: :class:`~Orange.classification.Classifier` - """ - def __init__(self, classifiers, meta_classifier, **kwds): - self.classifiers = classifiers - self.meta_classifier = meta_classifier - self.domain = Orange.data.Domain(self.meta_classifier.domain.features, False) - self.__dict__.update(kwds) - - def __call__(self, instance, resultType=Orange.core.GetValue): - ps = Orange.data.Instance(self.domain, [p for cl in self.classifiers for p in list(cl(instance, Orange.core.GetProbabilities))[:-1]]) - return self.meta_classifier(ps, resultType) Index: Orange/feature/scoring.py =================================================================== --- Orange/feature/scoring.py (revision 10524) +++ Orange/feature/scoring.py (revision 10541) @@ -1,14 +1,16 @@ -import Orange.core as orange -import Orange.misc - -from orange import MeasureAttribute as Score -from orange import MeasureAttributeFromProbabilities as ScoreFromProbabilities -from orange import MeasureAttribute_info as InfoGain -from orange import MeasureAttribute_gainRatio as GainRatio -from orange import MeasureAttribute_gini as Gini -from orange import MeasureAttribute_relevance as Relevance -from orange import MeasureAttribute_cost as Cost -from orange import MeasureAttribute_relief as Relief -from orange import MeasureAttribute_MSE as MSE +from Orange import core, feature +from Orange.statistics import contingency, distribution + +from Orange.misc import deprecated_keywords, deprecated_members + +Score = core.MeasureAttribute +ScoreFromProbabilities = core.MeasureAttributeFromProbabilities +InfoGain = core.MeasureAttribute_info +GainRatio = core.MeasureAttribute_gainRatio +Gini = core.MeasureAttribute_gini +Relevance = core.MeasureAttribute_relevance +Cost = core.MeasureAttribute_cost +Relief = core.MeasureAttribute_relief +MSE = core.MeasureAttribute_MSE ###### @@ -31,8 +33,8 @@ :param data: a data table used to score features - :type data: Orange.data.Table + :type data: :obj:`~Orange.data.Table` :param weight: meta attribute that stores weights of instances - :type weight: Orange.feature.Descriptor + :type weight: :obj:`~Orange.feature.Descriptor` """ @@ -46,5 +48,5 @@ return [x[0] for x in measured] -OrderAttributes = Orange.misc.deprecated_members({ +OrderAttributes = deprecated_members({ "measure": "score", }, wrap_methods=[])(OrderAttributes) @@ -59,8 +61,8 @@ """ - @Orange.misc.deprecated_keywords({"aprioriDist": "apriori_dist"}) + @deprecated_keywords({"aprioriDist": "apriori_dist"}) def __new__(cls, attr=None, data=None, apriori_dist=None, weightID=None): self = Score.__new__(cls) - if attr != None and data != None: + if attr is not None and data is not None: #self.__init__(**argkw) return self.__call__(attr, data, apriori_dist, weightID) @@ -68,13 +70,13 @@ return self - @Orange.misc.deprecated_keywords({"aprioriDist": "apriori_dist"}) + @deprecated_keywords({"aprioriDist": "apriori_dist"}) def __call__(self, attr, data, apriori_dist=None, weightID=None): """Score the given feature. :param attr: feature to score - :type attr: Orange.feature.Descriptor + :type attr: :obj:`~Orange.feature.Descriptor` :param data: a data table used to score features - :type data: Orange.data.table + :type data: :obj:`~Orange.data.Table` :param apriori_dist: @@ -82,12 +84,12 @@ :param weightID: meta feature used to weight individual data instances - :type weightID: Orange.feature.Descriptor + :type weightID: :obj:`~Orange.feature.Descriptor` """ import numpy - from orngContingency import Entropy + from orngContingency import Entropy #TODO: Move to new hierarchy if attr in data.domain: # if we receive attr as string we have to convert to variable attr = data.domain[attr] - attrClassCont = orange.ContingencyAttrClass(attr, data) + attrClassCont = contingency.VarClass(attr, data) dist = [] for vals in attrClassCont.values(): @@ -116,8 +118,8 @@ """ - @Orange.misc.deprecated_keywords({"aprioriDist": "apriori_dist"}) + @deprecated_keywords({"aprioriDist": "apriori_dist"}) def __new__(cls, attr=None, data=None, apriori_dist=None, weightID=None): self = Score.__new__(cls) - if attr != None and data != None: + if attr is not None and data is not None: #self.__init__(**argkw) return self.__call__(attr, data, apriori_dist, weightID) @@ -125,13 +127,13 @@ return self - @Orange.misc.deprecated_keywords({"aprioriDist": "apriori_dist"}) + @deprecated_keywords({"aprioriDist": "apriori_dist"}) def __call__(self, attr, data, apriori_dist=None, weightID=None): """Score the given feature. :param attr: feature to score - :type attr: Orange.feature.Descriptor + :type attr: :obj:`~Orange.feature.Descriptor` :param data: a data table used to score the feature - :type data: Orange.data.table + :type data: :obj:`~Orange.data.Table` :param apriori_dist: @@ -139,9 +141,9 @@ :param weightID: meta feature used to weight individual data instances - :type weightID: Orange.feature.Descriptor + :type weightID: :obj:`~Orange.feature.Descriptor` """ - attrClassCont = orange.ContingencyAttrClass(attr, data) - classDist = orange.Distribution(data.domain.classVar, data).values() + attrClassCont = contingency.VarClass(attr, data) + classDist = distribution.Distribution(data.domain.classVar, data).values() nCls = len(classDist) nEx = len(data) @@ -177,5 +179,5 @@ -@Orange.misc.deprecated_keywords({"attrList": "attr_list", "attrMeasure": "attr_score", "removeUnusedValues": "remove_unused_values"}) +@deprecated_keywords({"attrList": "attr_list", "attrMeasure": "attr_score", "removeUnusedValues": "remove_unused_values"}) def merge_values(data, attr_list, attr_score, remove_unused_values = 1): import orngCI @@ -183,5 +185,5 @@ newData = data.select(attr_list + [data.domain.class_var]) newAttr = orngCI.FeatureByCartesianProduct(newData, attr_list)[0] - dist = orange.Distribution(newAttr, newData) + dist = distribution.Distribution(newAttr, newData) activeValues = [] for i in range(len(newAttr.values)): @@ -213,5 +215,5 @@ return newAttr - reducedAttr = orange.EnumVariable(newAttr.name, values = [newAttr.values[i] for i in activeValues]) + reducedAttr = feature.Discrete.EnumVariable(newAttr.name, values = [newAttr.values[i] for i in activeValues]) reducedAttr.get_value_from = newAttr.get_value_from reducedAttr.get_value_from.class_var = reducedAttr @@ -220,5 +222,5 @@ ###### # from orngFSS -@Orange.misc.deprecated_keywords({"measure": "score"}) +@deprecated_keywords({"measure": "score"}) def score_all(data, score=Relief(k=20, m=50)): """Assess the quality of features using the given measure and return @@ -226,11 +228,10 @@ :param data: data table should include a discrete class. - :type data: :obj:`Orange.data.Table` + :type data: :obj:`~Orange.data.Table` :param score: feature scoring function. Derived from :obj:`~Orange.feature.scoring.Score`. Defaults to :obj:`~Orange.feature.scoring.Relief` with k=20 and m=50. - :type measure: :obj:`~Orange.feature.scoring.Score` - :rtype: :obj:`list`; a sorted (by descending score) list of - tuples (feature name, score) + :type score: :obj:`~Orange.feature.scoring.Score` + :rtype: :obj:`list`; a sorted list of tuples (feature name, score) """ Index: Orange/feature/selection.py =================================================================== --- Orange/feature/selection.py (revision 10523) +++ Orange/feature/selection.py (revision 10172) @@ -1,5 +1,3 @@ __docformat__ = 'restructuredtext' - -from operator import itemgetter import Orange.core as orange @@ -19,7 +17,5 @@ """ - return [x[0] for x in \ - sorted(scores, key=itemgetter(1), reverse=True)[:n] - ] + return [x[0] for x in sorted(scores)[:n]] bestNAtts = best_n Index: Orange/misc/__init__.py =================================================================== --- Orange/misc/__init__.py (revision 10521) +++ Orange/misc/__init__.py (revision 10327) @@ -911,9 +911,5 @@ @wraps(base.__new__) def _orange__new_wrapped(cls, data=None, **kwargs): - if base == object: - self = base.__new__(cls) - else: - self = base.__new__(cls, **kwargs) - + self = base.__new__(cls, **kwargs) if data: self.__init__(**kwargs) Index: Orange/multilabel/br.py =================================================================== --- Orange/multilabel/br.py (revision 10502) +++ Orange/multilabel/br.py (revision 9994) @@ -75,6 +75,5 @@ def __call__(self, instances, weight_id = 0, **kwds): if not Orange.multilabel.is_multilabel(instances): - raise TypeError("The given data set is not a multi-label data set" - " with class values 0 and 1.") + raise TypeError("The given data set is not a multi-label data set.") for k in kwds.keys(): Index: Orange/multilabel/brknn.py =================================================================== --- Orange/multilabel/brknn.py (revision 10502) +++ Orange/multilabel/brknn.py (revision 9994) @@ -94,6 +94,5 @@ def __call__(self, instances, weight_id = 0, **kwds): if not Orange.multilabel.is_multilabel(instances): - raise TypeError("The given data set is not a multi-label data set" - " with class values 0 and 1.") + raise TypeError("The given data set is not a multi-label data set.") for k in kwds.keys(): Index: Orange/multilabel/lp.py =================================================================== --- Orange/multilabel/lp.py (revision 10502) +++ Orange/multilabel/lp.py (revision 9994) @@ -84,6 +84,5 @@ def __call__(self, instances, base_learner = None, weight_id = 0, **kwds): if not Orange.multilabel.is_multilabel(instances): - raise TypeError("The given data set is not a multi-label data set" - " with class values 0 and 1.") + raise TypeError("The given data set is not a multi-label data set.") self.__dict__.update(kwds) Index: Orange/multilabel/mlknn.py =================================================================== --- Orange/multilabel/mlknn.py (revision 10502) +++ Orange/multilabel/mlknn.py (revision 10417) @@ -143,6 +143,5 @@ def __call__(self, instances, weight_id = 0, **kwds): if not Orange.multilabel.is_multilabel(instances): - raise TypeError("The given data set is not a multi-label data set" - " with class values 0 and 1.") + raise TypeError("The given data set is not a multi-label data set.") self.__dict__.update(kwds) Index: Orange/orng/orngABML.py =================================================================== --- Orange/orng/orngABML.py (revision 10518) +++ Orange/orng/orngABML.py (revision 9737) @@ -9,4 +9,6 @@ import math +from Orange.classification.rules import create_dichotomous_class as createDichotomousClass +from Orange.classification.rules import ConvertClass # regular expressions # exppression for testing validity of a set of arguments: @@ -284,36 +286,4 @@ - -class ConvertClass: - """ Converting class variables into dichotomous class variable. """ - def __init__(self, classAtt, classValue, newClassAtt): - self.classAtt = classAtt - self.classValue = classValue - self.newClassAtt = newClassAtt - - def __call__(self, example, returnWhat): - if example[self.classAtt] == self.classValue: - return Orange.data.Value(self.newClassAtt, self.classValue + "_") - else: - return Orange.data.Value(self.newClassAtt, "not " + self.classValue) - - -def create_dichotomous_class(domain, att, value, negate, removeAtt=None): - # create new variable - newClass = Orange.feature.Discrete(att.name + "_", values=[str(value) + "_", "not " + str(value)]) - positive = Orange.data.Value(newClass, str(value) + "_") - negative = Orange.data.Value(newClass, "not " + str(value)) - newClass.getValueFrom = ConvertClass(att, str(value), newClass) - - att = [a for a in domain.attributes] - newDomain = Orange.data.Domain(att + [newClass]) - newDomain.addmetas(domain.getmetas()) - if negate == 1: - return (newDomain, negative) - else: - return (newDomain, positive) - - - def addErrors(test_data, classifier): """ Main task of this function is to add probabilistic errors to examples.""" Index: Orange/regression/lasso.py =================================================================== --- Orange/regression/lasso.py (revision 10535) +++ Orange/regression/lasso.py (revision 10314) @@ -1,2 +1,102 @@ +"""\ +############################ +Lasso regression (``lasso``) +############################ + +.. index:: regression + +.. _`Lasso regression. Regression shrinkage and selection via the lasso`: + http://www-stat.stanford.edu/~tibs/lasso/lasso.pdf + + +`The Lasso `_ is a shrinkage +and selection method for linear regression. It minimizes the usual sum of squared +errors, with a bound on the sum of the absolute values of the coefficients. + +To fit the regression parameters on housing data set use the following code: + +.. literalinclude:: code/lasso-example.py + :lines: 7,9,10,11 + +.. autoclass:: LassoRegressionLearner + :members: + +.. autoclass:: LassoRegression + :members: + + +.. autoclass:: LassoRegressionLearner + :members: + +.. autoclass:: LassoRegression + :members: + +Utility functions +----------------- + +.. autofunction:: center + +.. autofunction:: get_bootstrap_sample + +.. autofunction:: permute_responses + + +======== +Examples +======== + +To predict values of the response for the first five instances +use the code + +.. literalinclude:: code/lasso-example.py + :lines: 14,15 + +Output + +:: + + Actual: 24.00, predicted: 24.58 + Actual: 21.60, predicted: 23.30 + Actual: 34.70, predicted: 24.98 + Actual: 33.40, predicted: 24.78 + Actual: 36.20, predicted: 24.66 + +To see the fitted regression coefficients, print the model + +.. literalinclude:: code/lasso-example.py + :lines: 17 + +The output + +:: + + Variable Coeff Est Std Error p + Intercept 22.533 + CRIM -0.000 0.023 0.480 + INDUS -0.010 0.023 0.300 + RM 1.303 0.994 0.000 *** + AGE -0.002 0.000 0.320 + PTRATIO -0.191 0.209 0.050 . + LSTAT -0.126 0.105 0.000 *** + Signif. codes: 0 *** 0.001 ** 0.01 * 0.05 . 0.1 empty 1 + + + For 7 variables the regression coefficient equals 0: + ZN + CHAS + NOX + DIS + RAD + TAX + B + +shows that some of the regression coefficients are equal to 0. + + + + + +""" + import Orange import numpy Index: Orange/testing/regression/results_reference/selection-bayes.py.linux2.txt =================================================================== --- Orange/testing/regression/results_reference/selection-bayes.py.linux2.txt (revision 10290) +++ Orange/testing/regression/results_reference/selection-bayes.py.linux2.txt (revision 10290) @@ -0,0 +1,3 @@ +Learner CA +Naive Bayes 0.903 +with FSS 0.867 Index: Orange/testing/regression/results_reference/selection-best3.py.linux2.txt =================================================================== --- Orange/testing/regression/results_reference/selection-best3.py.linux2.txt (revision 10290) +++ Orange/testing/regression/results_reference/selection-best3.py.linux2.txt (revision 10290) @@ -0,0 +1,4 @@ +Best 3 features: +adoption-of-the-budget-resolution +aid-to-nicaraguan-contras +anti-satellite-test-ban Index: Orange/testing/regression/results_reference/selection-filtered-learner.py.linux2.txt =================================================================== --- Orange/testing/regression/results_reference/selection-filtered-learner.py.linux2.txt (revision 10290) +++ Orange/testing/regression/results_reference/selection-filtered-learner.py.linux2.txt (revision 10290) @@ -0,0 +1,6 @@ +Learner CA +bayes 0.903 +filtered 0.874 + +Number of times attributes were used in cross-validation: +10 x adoption-of-the-budget-resolution Index: Orange/testing/regression/results_tests_20/modules_abcn2-rules.py.txt =================================================================== --- Orange/testing/regression/results_tests_20/modules_abcn2-rules.py.txt (revision 10520) +++ Orange/testing/regression/results_tests_20/modules_abcn2-rules.py.txt (revision 9951) @@ -1,11 +1,9 @@ IF TRUE THEN survived=no<1490.000, 711.000> IF sex=['female'] THEN survived=yes<126.000, 344.000> +IF status=['third'] THEN survived=no<528.000, 178.000> IF status=['first'] THEN survived=yes<122.000, 203.000> -IF sex=['male'] THEN survived=no<1364.000, 367.000> IF status=['crew'] THEN survived=no<673.000, 212.000> -IF status=['third'] THEN survived=no<528.000, 178.000> IF age=['adult'] THEN survived=no<1438.000, 654.000> IF age=['child'] THEN survived=yes<52.000, 57.000> -IF status=['second'] THEN survived=yes<167.000, 118.000> IF status=['second'] AND age=['child'] THEN survived=yes<0.000, 24.000> IF sex=['female'] AND status=['first'] THEN survived=yes<4.000, 141.000> @@ -14,5 +12,4 @@ IF status=['first'] AND age=['child'] THEN survived=yes<0.000, 6.000> IF sex=['female'] AND age=['adult'] THEN survived=yes<109.000, 316.000> -IF status=['second'] AND sex=['male'] THEN survived=no<154.000, 25.000> IF sex=['male'] AND age=['adult'] THEN survived=no<1329.000, 338.000> IF status=['third'] AND age=['adult'] THEN survived=no<476.000, 151.000> @@ -30,6 +27,4 @@ IF status=['third'] THEN survived=no<528.000, 178.000> IF status=['crew'] THEN survived=no<673.000, 212.000> -IF sex=['male'] THEN survived=no<1364.000, 367.000> -IF status=['second'] AND sex=['male'] THEN survived=no<154.000, 25.000> IF status=['first'] THEN survived=yes<122.000, 203.000> IF sex=['female'] AND status=['first'] THEN survived=yes<4.000, 141.000> Index: Orange/testing/regression/results_tests_20/modules_fss1.py.linux2.txt =================================================================== --- Orange/testing/regression/results_tests_20/modules_fss1.py.linux2.txt (revision 10290) +++ Orange/testing/regression/results_tests_20/modules_fss1.py.linux2.txt (revision 10290) @@ -0,0 +1,9 @@ +Attribute scores for best three attributes: +0.613 physician-fee-freeze +0.255 el-salvador-aid +0.228 synfuels-corporation-cutback + +Best 3 attributes: +adoption-of-the-budget-resolution +aid-to-nicaraguan-contras +anti-satellite-test-ban Index: Orange/testing/regression/results_tests_20/modules_fss3.py.linux2.txt =================================================================== --- Orange/testing/regression/results_tests_20/modules_fss3.py.linux2.txt (revision 10290) +++ Orange/testing/regression/results_tests_20/modules_fss3.py.linux2.txt (revision 10290) @@ -0,0 +1,3 @@ +Learner CA +Naive Bayes 0.903 +with FSS 0.867 Index: Orange/testing/regression/results_tests_20/modules_fss4.py.linux2.txt =================================================================== --- Orange/testing/regression/results_tests_20/modules_fss4.py.linux2.txt (revision 10290) +++ Orange/testing/regression/results_tests_20/modules_fss4.py.linux2.txt (revision 10290) @@ -0,0 +1,6 @@ +Learner CA +bayes 0.903 +filtered 0.874 + +Number of times attributes were used in cross-validation: +10 x adoption-of-the-budget-resolution Index: Orange/testing/regression/results_tests_20/modules_fss5.py.linux2.txt =================================================================== --- Orange/testing/regression/results_tests_20/modules_fss5.py.linux2.txt (revision 10290) +++ Orange/testing/regression/results_tests_20/modules_fss5.py.linux2.txt (revision 10290) @@ -0,0 +1,17 @@ +Before feature subset selection: +0.098 workclass +0.056 marital-status +0.049 relationship +0.022 race +0.019 capital-gain +0.012 education-num +0.002 capital-loss +0.001 hours-per-week +-0.002 age +-0.006 native-country +-0.011 occupation +-0.014 education +-0.022 fnlwgt +-0.025 sex + +After feature subset selection with margin 0.010: Index: Orange/testing/regression/results_tests_20/modules_fss6.py.linux2.txt =================================================================== --- Orange/testing/regression/results_tests_20/modules_fss6.py.linux2.txt (revision 10290) +++ Orange/testing/regression/results_tests_20/modules_fss6.py.linux2.txt (revision 10290) @@ -0,0 +1,17 @@ +Before feature subset selection (14 attributes): +0.098 workclass +0.056 marital-status +0.049 relationship +0.022 race +0.019 capital-gain +0.012 education-num +0.002 capital-loss +0.001 hours-per-week +-0.002 age +-0.006 native-country +-0.011 occupation +-0.014 education +-0.022 fnlwgt +-0.025 sex + +After feature subset selection with margin 0.010 (0 attributes): Index: Orange/testing/regression/results_tutorial/fss6.py.linux2.txt =================================================================== --- Orange/testing/regression/results_tutorial/fss6.py.linux2.txt (revision 10291) +++ Orange/testing/regression/results_tutorial/fss6.py.linux2.txt (revision 10291) @@ -0,0 +1,17 @@ +Before feature subset selection (14 attributes): +0.098 workclass +0.056 marital-status +0.049 relationship +0.022 race +0.019 capital-gain +0.012 education-num +0.002 capital-loss +0.001 hours-per-week +-0.002 age +-0.006 native-country +-0.011 occupation +-0.014 education +-0.022 fnlwgt +-0.025 sex + +After feature subset selection with margin 0.010 (0 attributes): Index: range/testing/unit/tests/test_feature_selection.py =================================================================== --- Orange/testing/unit/tests/test_feature_selection.py (revision 10522) +++ (revision ) @@ -1,61 +1,0 @@ -import Orange.misc.testing as testing -from Orange.feature import selection, scoring -import Orange - -from operator import itemgetter - -try: - import unittest2 as unittest -except ImportError: - import unittest - -class TestSelection(unittest.TestCase): - def setUp(self): - self.score = Orange.feature.scoring.Gini() - self.data = Orange.data.Table("lenses") - - self.scores = scoring.score_all(self.data, self.score) - - def test_best_n(self): - sorted_scores = sorted(self.scores, key=itemgetter(1), - reverse=True) - # Test the descending order of scores - self.assertEqual(self.scores, sorted_scores) - - # best 3 scores - best_3 = map(itemgetter(0), sorted_scores[:3]) - - # test best_n function - self.assertEqual(selection.best_n(self.scores, 3), best_3) - - self.assertTrue(len(selection.best_n(self.scores, 3)) == 3) - - # all returned values should be strings. - self.assertTrue(all(isinstance(item, basestring) for item in \ - selection.best_n(self.scores, 3))) - - new_data = selection.select_best_n(self.data, self.scores, 3) - self.assertEqual(best_3, [a.name for a in new_data.domain.attributes]) - self.assertEqual(new_data.domain.class_var, self.data.domain.class_var) - - def test_above_threashold(self): - threshold = self.scores[len(self.scores) / 2][1] - above = [item[0] for item in self.scores if item[1] > threshold] - - self.assertEqual(above, - selection.above_threshold(self.scores, threshold) - ) - - new_data = selection.select_above_threshold(self.data, - self.scores, threshold) - self.assertEqual(above, [a.name for a in new_data.domain.attributes]) - self.assertEqual(new_data.domain.class_var, self.data.domain.class_var) - - - - - - -if __name__ == "__main__": - unittest.main() - Index: docs/reference/rst/Orange.ensemble.rst =================================================================== --- docs/reference/rst/Orange.ensemble.rst (revision 10540) +++ docs/reference/rst/Orange.ensemble.rst (revision 9372) @@ -3,230 +3,4 @@ ################################## -.. index:: ensemble - -`Ensembles `_ use -multiple models to improve prediction performance. The module -implements a number of popular approaches, including bagging, -boosting, stacking and forest trees. Most of these are available both -for classification and regression with exception of stacking, which -with present implementation supports classification only. - -******* -Bagging -******* - -.. index:: bagging -.. index:: - single: ensemble; ensemble - -.. autoclass:: Orange.ensemble.bagging.BaggedLearner - :members: - :show-inheritance: - -.. autoclass:: Orange.ensemble.bagging.BaggedClassifier - :members: - :show-inheritance: - -******** -Boosting -******** - -.. index:: boosting -.. index:: - single: ensemble; boosting - - -.. autoclass:: Orange.ensemble.boosting.BoostedLearner - :members: - :show-inheritance: - -.. autoclass:: Orange.ensemble.boosting.BoostedClassifier - :members: - :show-inheritance: - -Example -======= - -The following script fits classification models by boosting and -bagging on Lymphography data set with TreeLearner and post-pruning as -a base learner. Classification accuracy of the methods is estimated by -10-fold cross validation (:download:`ensemble.py `): - -.. literalinclude:: code/ensemble.py - :lines: 7- - -Running this script demonstrates some benefit of boosting and bagging -over the baseline learner:: - - Classification Accuracy: - tree: 0.764 - boosted tree: 0.770 - bagged tree: 0.790 - -******** -Stacking -******** - -.. index:: stacking -.. index:: - single: ensemble; stacking - - -.. autoclass:: Orange.ensemble.stacking.StackedClassificationLearner - :members: - :show-inheritance: - -.. autoclass:: Orange.ensemble.stacking.StackedClassifier - :members: - :show-inheritance: - -Example -======= - -Stacking often produces classifiers that are more predictive than -individual classifiers in the ensemble. This effect is illustrated by -a script that combines four different classification -algorithms (:download:`ensemble-stacking.py `): - -.. literalinclude:: code/ensemble-stacking.py - :lines: 3- - -The benefits of stacking on this particular data set are -substantial (numbers show classification accuracy):: - - stacking: 0.934 - bayes: 0.858 - tree: 0.688 - lr: 0.764 - knn: 0.830 - -************* -Random Forest -************* - -.. index:: random forest -.. index:: - single: ensemble; random forest - -.. autoclass:: Orange.ensemble.forest.RandomForestLearner - :members: - :show-inheritance: - -.. autoclass:: Orange.ensemble.forest.RandomForestClassifier - :members: - :show-inheritance: - - -Example -======== - -The following script assembles a random forest learner and compares it -to a tree learner on a liver disorder (bupa) and housing data sets. - -:download:`ensemble-forest.py ` - -.. literalinclude:: code/ensemble-forest.py - :lines: 7- - -Notice that our forest contains 50 trees. Learners are compared through -3-fold cross validation:: - - Classification: bupa.tab - Learner CA Brier AUC - tree 0.586 0.829 0.575 - forest 0.710 0.392 0.752 - Regression: housing.tab - Learner MSE RSE R2 - tree 23.708 0.281 0.719 - forest 11.988 0.142 0.858 - -Perhaps the sole purpose of the following example is to show how to -access the individual classifiers once they are assembled into the -forest, and to show how we can assemble a tree learner to be used in -random forests. In the following example the best feature for decision -nodes is selected among three randomly chosen features, and maxDepth -and minExamples are both set to 5. - -:download:`ensemble-forest2.py ` - -.. literalinclude:: code/ensemble-forest2.py - :lines: 7- - -Running the above code would report on sizes (number of nodes) of the tree -in a constructed random forest. - - -Feature scoring -=============== - -L. Breiman (2001) suggested the possibility of using random forests as a -non-myopic measure of feature importance. - -The assessment of feature relevance with random forests is based on the -idea that randomly changing the value of an important feature greatly -affects instance's classification, while changing the value of an -unimportant feature does not affect it much. Implemented algorithm -accumulates feature scores over given number of trees. Importance of -all features for a single tree are computed as: correctly classified -OOB instances minus correctly classified OOB instances when the feature is -randomly shuffled. The accumulated feature scores are divided by the -number of used trees and multiplied by 100 before they are returned. - -.. autoclass:: Orange.ensemble.forest.ScoreFeature - :members: - -Computation of feature importance with random forests is rather slow -and importances for all features need to be computes -simultaneously. When it is called to compute a quality of certain -feature, it computes qualities for all features in the dataset. When -called again, it uses the stored results if the domain is still the -same and the data table has not changed (this is done by checking the -data table's version and is not foolproof; it will not detect if you -change values of existing instances, but will notice adding and -removing instances; see the page on :class:`Orange.data.Table` for -details). - -:download:`ensemble-forest-measure.py ` - -.. literalinclude:: code/ensemble-forest-measure.py - :lines: 7- - -The output of the above script is:: - - DATA:iris.tab - - first: 3.91, second: 0.38 - - different random seed - first: 3.39, second: 0.46 - - All importances: - sepal length: 3.39 - sepal width: 0.46 - petal length: 30.15 - petal width: 31.98 - -References ----------- - -* L Breiman. Bagging Predictors. `Technical report No. 421 - `_. University - of California, Berkeley, 1994. -* Y Freund, RE Schapire. `Experiments with a New Boosting Algorithm - `_. Machine - Learning: Proceedings of the Thirteenth International Conference - (ICML'96), 1996. -* JR Quinlan. `Boosting, bagging, and C4.5 - `_ . In Proc. of 13th - National Conference on Artificial Intelligence - (AAAI'96). pp. 725-730, 1996. -* L Brieman. `Random Forests - `_. Machine - Learning, 45, 5-32, 2001. -* M Robnik-Sikonja. `Improving Random Forests - `_. In - Proc. of European Conference on Machine Learning (ECML 2004), - pp. 359-370, 2004. - .. automodule:: Orange.ensemble Index: docs/reference/rst/Orange.regression.lasso.rst =================================================================== --- docs/reference/rst/Orange.regression.lasso.rst (revision 10536) +++ docs/reference/rst/Orange.regression.lasso.rst (revision 9372) @@ -1,95 +1,1 @@ -############################ -Lasso regression (``lasso``) -############################ - .. automodule:: Orange.regression.lasso - -.. index:: regression - -.. _`Lasso regression. Regression shrinkage and selection via the lasso`: - http://www-stat.stanford.edu/~tibs/lasso/lasso.pdf - - -`The Lasso `_ is a shrinkage -and selection method for linear regression. It minimizes the usual sum of squared -errors, with a bound on the sum of the absolute values of the coefficients. - -To fit the regression parameters on housing data set use the following code: - -.. literalinclude:: code/lasso-example.py - :lines: 9,10,11 - -.. autoclass:: LassoRegressionLearner - :members: - -.. autoclass:: LassoRegression - :members: - - -.. autoclass:: LassoRegressionLearner - :members: - -.. autoclass:: LassoRegression - :members: - -Utility functions ------------------ - -.. autofunction:: center - -.. autofunction:: get_bootstrap_sample - -.. autofunction:: permute_responses - - -======== -Examples -======== - -To predict values of the response for the first five instances -use the code - -.. literalinclude:: code/lasso-example.py - :lines: 14,15 - -Output - -:: - - Actual: 24.00, predicted: 24.58 - Actual: 21.60, predicted: 23.30 - Actual: 34.70, predicted: 24.98 - Actual: 33.40, predicted: 24.78 - Actual: 36.20, predicted: 24.66 - -To see the fitted regression coefficients, print the model - -.. literalinclude:: code/lasso-example.py - :lines: 17 - -The output - -:: - - Variable Coeff Est Std Error p - Intercept 22.533 - CRIM -0.000 0.023 0.480 - INDUS -0.010 0.023 0.300 - RM 1.303 0.994 0.000 *** - AGE -0.002 0.000 0.320 - PTRATIO -0.191 0.209 0.050 . - LSTAT -0.126 0.105 0.000 *** - Signif. codes: 0 *** 0.001 ** 0.01 * 0.05 . 0.1 empty 1 - - - For 7 variables the regression coefficient equals 0: - ZN - CHAS - NOX - DIS - RAD - TAX - B - -shows that some of the regression coefficients are equal to 0. - Index: docs/reference/rst/Orange.regression.rst =================================================================== --- docs/reference/rst/Orange.regression.rst (revision 10537) +++ docs/reference/rst/Orange.regression.rst (revision 10396) @@ -3,10 +3,19 @@ ########################### -Orange implements a set of methods for regression modeling, that is, -where the outcome - dependent variable is real-valued: +Orange uses the term `classification` to also denote the +regression. For instance, the dependent variable is called a `class +variable` even when it is continuous, and models are generally called +classifiers. A part of the reason is that classification and +regression rely on the same set of basic classes. + +Please see the documentation on :doc:`Orange.classification` for +information on how to fit models in general. + +Orange contains a number of regression models which are listed below. .. toctree:: :maxdepth: 1 + Orange.regression.mean Orange.regression.linear Orange.regression.lasso @@ -14,16 +23,4 @@ Orange.regression.earth Orange.regression.tree - Orange.regression.mean - -Notice that the dependent variable is in this documentation and in the -implementation referred to as `class variable`. See also the documentation -on :doc:`Orange.classification` for information on how to fit models -and use them for prediction. - -************************* -Base class for regression -************************* - -All regression learners are inherited from `BaseRegressionLearner`. .. automodule:: Orange.regression.base Index: ocs/reference/rst/code/ensemble-stacking.py =================================================================== --- docs/reference/rst/code/ensemble-stacking.py (revision 10540) +++ (revision ) @@ -1,15 +1,0 @@ -import Orange - -data = Orange.data.Table("promoters") - -bayes = Orange.classification.bayes.NaiveLearner(name="bayes") -tree = Orange.classification.tree.SimpleTreeLearner(name="tree") -lin = Orange.classification.svm.LinearLearner(name="lr") -knn = Orange.classification.knn.kNNLearner(name="knn") - -base_learners = [bayes, tree, lin, knn] -stack = Orange.ensemble.stacking.StackedClassificationLearner(base_learners) - -learners = [stack, bayes, tree, lin, knn] -res = Orange.evaluation.testing.cross_validation(learners, data, 3) -print "\n".join(["%8s: %5.3f" % (l.name, r) for r, l in zip(Orange.evaluation.scoring.CA(res), learners)]) Index: install-scripts/createSnapshot.btm =================================================================== --- install-scripts/createSnapshot.btm (revision 10503) +++ install-scripts/createSnapshot.btm (revision 10275) @@ -9,4 +9,7 @@ set ADDON_TEXT_SNAPSHOT=orangeAddOn-text-snapshot-hg-%daystr set SOURCE_SNAPSHOT=orange-source-snapshot-hg-%daystr.zip + +rem # update source(s) to revision HEAD +cdd %TMPDIR rem # build core @@ -24,5 +27,5 @@ del /syeqtx sourcestopack -call hg clone . sourcestopack +call hg archive sourcestopack cd sourcestopack @@ -30,35 +33,18 @@ rem # build source distribution e:\Python27\python.exe setup.py sdist - rem # build msi file -rem # build_msi only handles StrictVersion so we change Orange\version.py -rem # and remove the .hg dir so the short_version is used -e:\Python27\python.exe -c"print open('Orange/version.py', 'rb').read().replace('if not release:', 'if not release and False:')" > version.py.1 -move /z version.py.1 Orange\version.py -del /syeqtx .hg - e:\Python27\python.exe setup.py bdist_msi -break_on_error - -rem # if no errors then publish on web (everything at the same time) -rem # remove any old files -rem # leave 10 latest versions. -e:\Python27\python.exe -c"import os,glob; [os.remove(f) for f in sorted(glob.glob('%DOWNLOADDIR\orange-win-snapshot-hg-*.exe'), reverse=True)[30:]]" -e:\Python27\python.exe -c"import os,glob; [os.remove(f) for f in sorted(glob.glob('%DOWNLOADDIR\orange-win-w-python-snapshot-hg-*.exe'), reverse=True)[30:]]" -e:\Python27\python.exe -c"import os,glob; [os.remove(f) for f in sorted(glob.glob('%DOWNLOADDIR\orangeAddOn-bioinformatics-snapshot-hg-*.exe'), reverse=True)[30:]]" -e:\Python27\python.exe -c"import os,glob; [os.remove(f) for f in sorted(glob.glob('%DOWNLOADDIR\orangeAddOn-text-snapshot-hg-*.exe'), reverse=True)[30:]]" - -e:\Python27\python.exe -c"import os,glob; [os.remove(f) for f in sorted(glob.glob('%DOWNLOADDIR\orange-source-snapshot-hg-*.zip'), reverse=True)[10:]]" -e:\Python27\python.exe -c"import os,glob; [os.remove(f) for f in sorted(glob.glob('%DOWNLOADDIR\orange-win-snapshot-hg-*.exe'), reverse=True)[10:]]" -e:\Python27\python.exe -c"import os,glob; [os.remove(f) for f in sorted(glob.glob('%DOWNLOADDIR\orange-win-snapshot-hg-*.msi'), reverse=True)[10:]]" - -rem except (%DOWNLOADDIR\orange*-snapshot-%daystr-*.exe) del %DOWNLOADDIR\orange*-snapshot-????-??-??-*.exe - -rem # publish - +rem del %DOWNLOADDIR\orange-source-snapshot-????-??-??.zip move /z dist\Orange-?.*.zip %DOWNLOADDIR\%SOURCE_SNAPSHOT move /z dist\Orange-?.*.msi %DOWNLOADDIR\%WIN_SNAPSHOT.msi + +rem # if no errors then publish on web (everything at the same time) +rem # remove any old files + +except (%DOWNLOADDIR\orange*-snapshot-%daystr-*.exe) del %DOWNLOADDIR\orange*-snapshot-????-??-??-*.exe + +rem # publish cdd %TMPDIR move /z *.exe %DOWNLOADDIR Index: install-scripts/mac/build-source.sh =================================================================== --- install-scripts/mac/build-source.sh (revision 10508) +++ install-scripts/mac/build-source.sh (revision 10445) @@ -35,6 +35,6 @@ # Build the source distribution -BUILD_TAG=`hg parent --template=".dev-r{rev}-{node|short}"` -python setup.py egg_info --tag-build=$BUILD_TAG sdist +hg parent --template="{latesttag}{latesttagdistance}.dev-{node|short}" > VERSION.txt +python setup.py sdist # Copy the source an egg info to workdir Index: install-scripts/mac/bundle-build-hg.sh =================================================================== --- install-scripts/mac/bundle-build-hg.sh (revision 10504) +++ install-scripts/mac/bundle-build-hg.sh (revision 10488) @@ -20,6 +20,5 @@ echo "Preaparing the bundle template" -TEMPLATE_VERSION=`curl --silent http://orange.biolab.si/download/bundle-templates/CURRENT.txt` -curl --silent http://orange.biolab.si/download/bundle-templates/Orange-template-${TEMPLATE_VERSION}.tar.gz | tar -xz -C $WORK_DIR +curl --silent http://orange.biolab.si/download/bundle-templates/Orange-template-2.0a2.tar.gz | tar -xz -C $WORK_DIR #svn export --non-interactive http://orange.biolab.si/svn/orange/externals/trunk/install-scripts/mac/bundle/ $TMP_BUNDLE_DIR Index: install-scripts/mac/dailyrun.sh =================================================================== --- install-scripts/mac/dailyrun.sh (revision 10527) +++ install-scripts/mac/dailyrun.sh (revision 10345) @@ -1,9 +1,6 @@ -#!/bin/bash -e +#!/bin/bash # # Should be run as: sudo ./dailyrun.sh # - -#FORCE=true -#LOCAL=true test -r /sw/bin/init.sh && . /sw/bin/init.sh @@ -11,49 +8,29 @@ export PATH=$HOME/bin:$PATH -WORK_DIR=/private/tmp/repos +STABLE_REVISION_1=`svn info --non-interactive http://orange.biolab.si/svn/orange/branches/ver1.0/ | grep 'Last Changed Rev:' | cut -d ' ' -f 4` +# svn info does not return proper exit status on an error so we check it this way +[ "$STABLE_REVISION_1" ] || exit 1 +STABLE_REVISION_2=`svn info --non-interactive http://orange.biolab.si/svn/orange/externals/branches/ver1.0/ | grep 'Last Changed Rev:' | cut -d ' ' -f 4` +# svn info does not return proper exit status on an error so we check it this way +[ "$STABLE_REVISION_2" ] || exit 1 -if [ $LOCAL ]; then - PUBLISH_DIR=/private/tmp/download - mkdir -p $PUBLISH_DIR +if [[ $STABLE_REVISION_1 -gt $STABLE_REVISION_2 ]]; then + STABLE_REVISION=$STABLE_REVISION_1 else - PUBLISH_DIR=/Volumes/download + STABLE_REVISION=$STABLE_REVISION_2 fi -if [ ! -e $WORK_DIR ]; then - mkdir -p $WORK_DIR +DAILY_REVISION_1=`svn info --non-interactive http://orange.biolab.si/svn/orange/trunk/ | grep 'Last Changed Rev:' | cut -d ' ' -f 4` +# svn info does not return proper exit status on an error so we check it this way +[ "$DAILY_REVISION_1" ] || exit 1 +DAILY_REVISION_2=`svn info --non-interactive http://orange.biolab.si/svn/orange/externals/trunk/ | grep 'Last Changed Rev:' | cut -d ' ' -f 4` +# svn info does not return proper exit status on an error so we check it this way +[ "$DAILY_REVISION_2" ] || exit 1 + +if [[ $DAILY_REVISION_1 -gt $DAILY_REVISION_2 ]]; then + DAILY_REVISION=$DAILY_REVISION_1 +else + DAILY_REVISION=$DAILY_REVISION_2 fi - -SOURCE_LOG=/private/tmp/sources-daily-build.log - -# Build source packages -./build-source.sh https://bitbucket.org/biolab/orange orange tip $WORK_DIR Orange > $SOURCE_LOG 2>&1 -EXIT_VALUE1=$? -./build-source.sh https://bitbucket.org/biolab/orange-addon-bioinformatics bioinformatics tip $WORK_DIR Orange-Bioinformatics >> $SOURCE_LOG 2>&1 -EXIT_VALUE2=$? -./build-source.sh https://bitbucket.org/biolab/orange-addon-text text tip $WORK_DIR Orange-Text-Mining >> $SOURCE_LOG 2>&1 -EXIT_VALUE3=$? - -echo "Orange (sources) [$EXIT_VALUE1 $EXIT_VALUE2 $EXIT_VALUE3]" > "/Volumes/download/buildLogs/osx/source-daily-build-hg.log" -date >> "/Volumes/download/buildLogs/osx/source-daily-build-hg.log" -cat $SOURCE_LOG > "/Volumes/download/buildLogs/osx/source-daily-build-hg.log" -(($EXIT_VALUE1 + $EXIT_VALUE2 + $EXIT_VALUE3)) && echo "Daily sources failed" - -# Get versions from PKG-INFO files -ORANGE_VERSION=`grep "^Version:" $WORK_DIR/Orange.egg-info/PKG-INFO | cut -d " " -f 2` -BIOINFORMATICS_VERSION=`grep "^Version:" $WORK_DIR/Orange_Bioinformatics.egg-info/PKG-INFO | cut -d " " -f 2` -TEXT_VERSION=`grep "^Version:" $WORK_DIR/Orange_Text_Mining.egg-info/PKG-INFO | cut -d " " -f 2` - - -# Source filenames -ORANGE_SOURCE="Orange-${ORANGE_VERSION}.tar.gz" -BIOINFORMATICS_SOURCE="Orange-Bioinformatics-${BIOINFORMATICS_VERSION}.tar.gz" -TEXT_SOURCE="Orange-Text-Mining-${TEXT_VERSION}.tar.gz" - - -# Get source packages md5 checksum -ORANGE_SOURCE_MD5=`md5 -q $WORK_DIR/$ORANGE_SOURCE` -BIOINFORMATICS_SOURCE_MD5=`md5 -q $WORK_DIR/$BIOINFORMATICS_SOURCE` -TEXT_SOURCE_MD5=`md5 -q $WORK_DIR/$TEXT_SOURCE` - MAC_VERSION=`sw_vers -productVersion | cut -d '.' -f 2` @@ -62,85 +39,11 @@ defaults write com.apple.desktopservices DSDontWriteNetworkStores true -if [ ! $LOCAL ]; then - /Users/ailabc/mount-dirs.sh || { echo "Mounting failed." ; exit 1 ; } -fi +/Users/ailabc/mount-dirs.sh || { echo "Mounting failed." ; exit 1 ; } -# Base url for sources -if [ $LOCAL ]; then - BASE_URL="file://$PUBLISH_DIR/sources" -else - BASE_URL="http://orange.biolab.si/download/sources" -fi +## Daily build from hg +/Users/ailabc/bundle-daily-build-hg.sh &> /private/tmp/bundle-daily-build.log +EXIT_VALUE=$? -# Base dir for sources -SOURCES_DIR=$PUBLISH_DIR/sources - - -# Publish sources - -if [ ! -e $SOURCES_DIR ]; then - mkdir -p $SOURCES_DIR -fi - -if [[ ! -e $SOURCES_DIR/$ORANGE_SOURCE || $FORCE ]]; then - cp $WORK_DIR/$ORANGE_SOURCE $SOURCES_DIR/$ORANGE_SOURCE - NEW_ORANGE=1 -fi - -if [[ ! -e $SOURCES_DIR/BIOINFORMATICS_SOURCE || $FORCE ]]; then - cp $WORK_DIR/$BIOINFORMATICS_SOURCE $SOURCES_DIR/$BIOINFORMATICS_SOURCE - NEW_BIOINFORMATICS=1 -fi - -if [[ ! -e $SOURCES_DIR/TEXT_SOURCE || $FORCE ]]; then - cp $WORK_DIR/$TEXT_SOURCE $SOURCES_DIR/$TEXT_SOURCE - NEW_TEXT=1 -fi - -FINK_ROOT=/sw - -# Update the local finkinfo -# Local info files will be copied to biolab/main/finkinfo in fink-daily-build-packages.sh -FINK_INFO_DIR="$FINK_ROOT/fink/dists/local/main/finkinfo" - -if [ ! -e $FINK_INFO_DIR ]; then - mkdir -p $FINK_INFO_DIR -fi - -# Directory where fink .info templates are -FINK_TEMPLATES=$WORK_DIR/orange/install-scripts/mac/fink - -FINK_LOG=/private/tmp/bundle-daily-build.log -echo "" > $FINK_LOG - -if [[ $NEW_ORANGE || $FORCE ]]; then - FINK_ORANGE_SOURCE_TEMPLATE="Orange-%v.tar.gz" - ./fink-register-info.sh "$FINK_TEMPLATES/orange-gui-hg-py.info" $BASE_URL/$FINK_ORANGE_SOURCE_TEMPLATE $ORANGE_SOURCE_MD5 $ORANGE_VERSION $FINK_INFO_DIR/orange-gui-hg-py.info >> $FINK_LOG 2>&1 -fi - -if [[ $NEW_BIOINFORMATICS || $FORCE ]]; then - FINK_BIOINFORMATICS_SOURCE_TEMPLATE="Orange-Bioinformatics-%v.tar.gz" - ./fink-register-info.sh "$FINK_TEMPLATES/orange-bioinformatics-gui-hg-py.info" $BASE_URL/$FINK_BIOINFORMATICS_SOURCE_TEMPLATE $BIOINFORMATICS_SOURCE_MD5 $BIOINFORMATICS_VERSION $FINK_INFO_DIR/orange-bioinformatics-gui-hg-py.info >> $FINK_LOG 2>&1 -fi - -if [[ $NEW_TEXT || $FORCE ]]; then - FINK_TEXT_SOURCE_TEMPLATE="Orange-Text-Mining-%v.tar.gz" - ./fink-register-info.sh "$FINK_TEMPLATES/orange-text-gui-hg-py.info" $BASE_URL/$FINK_TEXT_SOURCE_TEMPLATE $TEXT_SOURCE_MD5 $TEXT_VERSION $FINK_INFO_DIR/orange-text-gui-hg-py.info >> $FINK_LOG 2>&1 -fi - -if [ ! $LOCAL ]; then - /Users/ailabc/mount-dirs.sh || { echo "Mounting failed." ; exit 1 ; } -fi - - -## Daily bundle build from hg -if [[ $NEW_ORANGE || $NEW_BIOINFORMATICS || $NEW_TEXT || $FORCE ]]; then - /Users/ailabc/bundle-daily-build-hg.sh &> /private/tmp/bundle-daily-build.log - EXIT_VALUE=$? -fi - -if [ ! $LOCAL ]; then - /Users/ailabc/mount-dirs.sh || { echo "Mounting failed." ; exit 1 ; } -fi +/Users/ailabc/mount-dirs.sh || { echo "Mounting failed." ; exit 1 ; } echo "Orange (bundle $MAC_VERSION from hg) [$EXIT_VALUE]" > "/Volumes/download/buildLogs/osx/bundle-$MAC_VERSION-daily-build-hg.log" @@ -150,16 +53,32 @@ -## daily fink build +#/Users/ailabc/bundle-daily-build.sh $STABLE_REVISION $DAILY_REVISION &> /private/tmp/bundle-daily-build.log +#EXIT_VALUE=$? +# +#/Users/ailabc/mount-dirs.sh || { echo "Mounting failed." ; exit 1 ; } +# +#echo "Orange (bundle $MAC_VERSION) [$EXIT_VALUE]" > "/Volumes/download/buildLogs/osx/bundle-$MAC_VERSION-daily-build.log" +#date >> "/Volumes/download/buildLogs/osx/bundle-$MAC_VERSION-daily-build.log" +#cat /private/tmp/bundle-daily-build.log >> "/Volumes/download/buildLogs/osx/bundle-$MAC_VERSION-daily-build.log" +#(($EXIT_VALUE)) && echo "Running bundle-daily-build.sh failed" +# +#/Users/ailabc/bundle-64bit-daily-build.sh $DAILY_REVISION &> /private/tmp/bundle-64bit-daily-build.log +#EXIT_VALUE=$? -/Users/ailabc/fink-daily-build-packages.sh &> /private/tmp/fink-daily-build-packages.log +#/Users/ailabc/mount-dirs.sh || { echo "Mounting failed." ; exit 1 ; } +# +#echo "Orange (bundle $MAC_VERSION 64bit) [$EXIT_VALUE]" > "/Volumes/download/buildLogs/osx/bundle-$MAC_VERSION-64bit-daily-build.log" +#date >> "/Volumes/download/buildLogs/osx/bundle-$MAC_VERSION-64bit-daily-build.log" +#cat /private/tmp/bundle-64bit-daily-build.log >> "/Volumes/download/buildLogs/osx/bundle-$MAC_VERSION-64bit-daily-build.log" +#(($EXIT_VALUE)) && echo "Running bundle-64bit-daily-build.sh failed" + +/Users/ailabc/fink-daily-build.sh $STABLE_REVISION $DAILY_REVISION &> /private/tmp/fink-daily-build.log EXIT_VALUE=$? -if [ ! $LOCAL ]; then - /Users/ailabc/mount-dirs.sh || { echo "Mounting failed." ; exit 1 ; } -fi +/Users/ailabc/mount-dirs.sh || { echo "Mounting failed." ; exit 1 ; } echo "Orange (fink $MAC_VERSION $ARCH) [$EXIT_VALUE]" > "/Volumes/download/buildLogs/osx/fink-$MAC_VERSION-$ARCH-daily-build.log" date >> "/Volumes/download/buildLogs/osx/fink-$MAC_VERSION-$ARCH-daily-build.log" -cat /private/tmp/fink-daily-build-packages.log >> "/Volumes/download/buildLogs/osx/fink-$MAC_VERSION-$ARCH-daily-build.log" +cat /private/tmp/fink-daily-build.log >> "/Volumes/download/buildLogs/osx/fink-$MAC_VERSION-$ARCH-daily-build.log" (($EXIT_VALUE)) && echo "Running fink-daily-build.sh failed" Index: nstall-scripts/mac/fink-daily-build-packages.sh =================================================================== --- install-scripts/mac/fink-daily-build-packages.sh (revision 10500) +++ (revision ) @@ -1,256 +1,0 @@ -#!/bin/bash -e -# -# Run daily fink build -# - -# Daily orange packages to build -DAILY_PACKAGES="orange-hg-py26 orange-gui-hg-py26 orange-bioinformatics-hg-py26 orange-bioinformatics-gui-hg-py26 orange-text-hg-py26 orange-text-gui-hg-py26 orange-hg-py27 orange-gui-hg-py27 orange-bioinformatics-hg-py27 orange-bioinformatics-gui-hg-py27 orange-text-hg-py27 orange-text-gui-hg-py27" - -# Packages which, when installing, want special confirmation from the user -# We keep those packages installed all the time -SPECIAL_PACKAGES="passwd xinitrc" - -# A list of packages (dependencies) from which user can choose upon installing our packages -# We would like to build all those so that it does not need to compile anything whichever packages he or she chooses -# The problem is that they are often mutually conflicting so we cannot have them simply installed (so that update-all -# would update them) but have to build them explicitly -OTHER_PACKAGES="ghostscript ghostscript-esp ghostscript6 ghostscript-nox ghostscript6-nox gnuplot gnuplot-nox gnuplot-nogtk tetex-base tetex-nox-base texlive-nox-base texlive-base tetex-texmf texlive-texmf" - - -# Miscellaneous extra packages which are maybe not really needed for Orange but are useful for CS research -EXTRA_PACKAGES="fuse gcc42 gcc43 gcc44 gnuplot gnuplot-nox gnuplot-nogtk db48 db48-aes git imagemagick-nox rrdtool maxima nmap wireshark openssl pstree python26 python27 python3 rdiff-backup svn swi-prolog lynx links w3m elinks matplotlib-py26 matplotlib-py27 mercurial-py26 mercurial-py27" - -FINK_ARGS="--yes --build-as-nobody" -FINK_SELFUPDATE_ARGS="--yes" -APT_ARGS="--assume-yes" - -# Path to Fink root -FINK_ROOT=/sw - -# Repo dir -REPO_DIR=/private/tmp/repos - -ARCH=`perl -MFink::FinkVersion -e 'print Fink::FinkVersion::get_arch'` - -# Sets error handler -trap "echo \"Script failed\"" ERR - -((`id -u` == 0)) || { echo "Must run as root user (use sudo)."; exit 1; } - -test -r $FINK_ROOT/bin/init.sh || { echo "Fink cannot be found." exit 2; } - -[ -e /Volumes/fink/ ] || { echo "/Volumes/fink/ not mounted."; exit 3; } - -# Configures environment for Fink -. $FINK_ROOT/bin/init.sh - -if ! grep '^Trees:' $FINK_ROOT/etc/fink.conf | grep -q 'unstable/main' && grep '^SelfUpdateMethod:' $FINK_ROOT/etc/fink.conf | grep -q 'point'; then - echo "Fink does not seem to use unstable Fink packages tree with rsync or CVS updating." - exit 5 -fi - -if [ ! -x /usr/bin/xcodebuild ]; then - echo "It seems Xcode is not installed on a system." - exit 6 -fi - -MAC_VERSION=`sw_vers -productVersion | cut -d '.' -f 2` -if [[ "$MAC_VERSION" -ne 5 && "$MAC_VERSION" -ne 6 ]]; then - echo "It seems system is not Mac OS X version 10.5 or 10.6." - exit 7 -fi - -if [ ! "`/usr/X11/bin/X -version 2>&1 | grep '^X.Org X Server' | grep -E -o '[0-9]+\.[0-9]+\.[0-9]+' | cut -d '.' -f 2`" -gt "3" ]; then - echo "It seems X11 version 2.3.0 or later is not installed on a system." - exit 8 -fi - -echo "Preparing local biolab Fink info files repository." -mkdir -p $FINK_ROOT/fink/dists/biolab/main/finkinfo/ -rm -f $FINK_ROOT/fink/dists/biolab/main/finkinfo/* - -echo "Updating local biolab Fink info files repository." -curl "http://orange.biolab.si/fink/dists/10.$MAC_VERSION/main/finkinfo/all.tgz" --output $FINK_ROOT/fink/dists/biolab/main/finkinfo/all.tgz -tar -xzf $FINK_ROOT/fink/dists/biolab/main/finkinfo/all.tgz -C $FINK_ROOT/fink/dists/biolab/main/finkinfo/ -rm -f $FINK_ROOT/fink/dists/biolab/main/finkinfo/all.tgz - -# Copy info files from local/main/finkinfo -echo "Updating new fink info files." -mv $FINK_ROOT/fink/dists/local/main/finkinfo/*.info $FINK_ROOT/fink/dists/biolab/main/finkinfo/ - -if ! grep '^Trees:' $FINK_ROOT/etc/fink.conf | grep -q 'biolab/main'; then - echo "Adding local biolab Fink info files repository to Fink configuration." - perl -p -i -l -e '$_ = "$_ biolab/main" if /^Trees/' $FINK_ROOT/etc/fink.conf -fi - -# Adds our binary repository to local Fink (APT) configuration -if ! grep -q "deb http://orange.biolab.si/fink 10.$MAC_VERSION main" $FINK_ROOT/etc/apt/sources.list; then - echo "Adding biolab Fink binary packages repository to Fink configuration." - echo "deb http://orange.biolab.si/fink 10.$MAC_VERSION main" >> $FINK_ROOT/etc/apt/sources.list -fi - -if [ ! -e $FINK_ROOT/etc/apt/apt.conf.d/daily-build ]; then - echo "Configuring apt-get to assume yes to all questions." - echo 'APT::Get::Assume-Yes "true";' > $FINK_ROOT/etc/apt/apt.conf.d/daily-build -fi - -# Configures any pending packages from possible interrupted past sessions -dpkg --configure -a - -# Gets all official Fink package info files -echo "Updating installed Fink packages." -fink $FINK_SELFUPDATE_ARGS selfupdate --method=rsync -fink $FINK_ARGS scanpackages - -# Updates everything (probably by compiling new packages) -fink $FINK_ARGS update-all - -# Installs special packages (if they are not already installed) -yes | fink $FINK_ARGS install $SPECIAL_PACKAGES - -# Removes possiblly installed packages which we want built -fink $FINK_ARGS purge --recursive $DAILY_PACKAGES $OTHER_PACKAGES $EXTRA_PACKAGES -# Sometimes Fink and APT are not in sync so we remove packages also directly -for package in $DAILY_PACKAGES $OTHER_PACKAGES $EXTRA_PACKAGES ; do - echo $package "purge" | dpkg --set-selections -done -apt-get $APT_ARGS dselect-upgrade - -# Stores current packages status -dpkg --get-selections '*' > /tmp/dpkg-selections.list - -for package in $OTHER_PACKAGES ; do - # Restores intitial packages status - dpkg --get-selections '*' | cut -f 1 | xargs -n 1 -J % echo % purge | dpkg --set-selections - dpkg --set-selections < /tmp/dpkg-selections.list - apt-get $APT_ARGS dselect-upgrade - - # Builds a package if it has not been rebuilt already (for example, as a dependency) - # We install it and not just build it because installation does not build package if it already exists as a binary package - echo "Specially building package $package." - fink $FINK_ARGS install $package -done - -for package in $EXTRA_PACKAGES ; do - if fink $FINK_ARGS describe $package > /dev/null ; then - # Restores intitial packages status - dpkg --get-selections '*' | cut -f 1 | xargs -n 1 -J % echo % purge | dpkg --set-selections - dpkg --set-selections < /tmp/dpkg-selections.list - apt-get $APT_ARGS dselect-upgrade - - # Builds a package if it has not been rebuilt already (for example, as a dependency) - # We install it and not just build it because installation does not build package if it already exists as a binary package - echo "Specially building extra package $package." - fink $FINK_ARGS install $package - else - echo "Not building extra package $package." - fi -done - -# We build our packages in "maintainer" mode - Fink makes tests and validates packages -for package in $DAILY_PACKAGES ; do - DEPS=`perl -MFink -MFink::PkgVersion -l -e "Fink::Package->require_packages(); map { map { /(\\S+)/; print \\$1 } @\\$_ } @{Fink::PkgVersion->match_package('$package')->get_depends(1, 0)};"` - - # First builds all dependencies normally (so that we are not checking for others' errors) - for deps in $DEPS ; do - # Restores intitial packages status - dpkg --get-selections '*' | cut -f 1 | xargs -n 1 -J % echo % purge | dpkg --set-selections - dpkg --set-selections < /tmp/dpkg-selections.list - apt-get $APT_ARGS dselect-upgrade - - # We install it and not just build it because installation does not build package if it already exists as a binary package - echo "Specially building package $package dependency $deps." - fink $FINK_ARGS install $deps - done - - # Restores intitial packages status - dpkg --get-selections '*' | cut -f 1 | xargs -n 1 -J % echo % purge | dpkg --set-selections - dpkg --set-selections < /tmp/dpkg-selections.list - apt-get $APT_ARGS dselect-upgrade - - # Then builds a package - # We can just build it as our packages have been probably cached if they have been already built - echo "Specially building, testing and validating package $package." - fink $FINK_ARGS --maintainer build $package -done - -echo "Restoring initial packages status." -dpkg --get-selections '*' | cut -f 1 | xargs -n 1 -J % echo % purge | dpkg --set-selections -dpkg --set-selections < /tmp/dpkg-selections.list -apt-get $APT_ARGS dselect-upgrade -rm -f /tmp/dpkg-selections.list - -# Cleans unncessary files (we cache them anyway in public repository) -echo "Cleaning." -fink $FINK_ARGS cleanup --all - -# TODO: Should be called only on a daily build server and not if building locally -/Users/ailabc/mount-dirs.sh - -echo "Preparing public biolab Fink info and binary files repository." -mkdir -p /Volumes/fink/dists/10.$MAC_VERSION/main/binary-darwin-$ARCH/ -chmod +rx /Volumes/fink/dists/10.$MAC_VERSION/main/binary-darwin-$ARCH/ -mkdir -p /Volumes/fink/dists/10.$MAC_VERSION/main/finkinfo/ -chmod +rx /Volumes/fink/dists/10.$MAC_VERSION/main/binary-darwin-$ARCH/ - -echo "Copying to repository all binary packages." -cp $FINK_ROOT/fink/debs/*.deb /Volumes/fink/dists/10.$MAC_VERSION/main/binary-darwin-$ARCH/ -if (shopt -s nullglob; f=($FINK_ROOT/var/cache/apt/archives/*.deb); ((${#f[@]}))); then - # We have to test if there are any deb files available as otherwise cp fails - cp $FINK_ROOT/var/cache/apt/archives/*.deb /Volumes/fink/dists/10.$MAC_VERSION/main/binary-darwin-$ARCH/ -fi - -cd /Volumes/fink/dists/10.$MAC_VERSION/main/binary-darwin-$ARCH/ - -echo "Fixing possible problems with binary packages filenames." -# Some packages include Fink epoch which uses colon as a delimiter and breaks package retrieval from the repository web server -# We remove epoch as it should not be there in the first place -perl -e ' -for (<*.deb>) { - if (m/^(.+)_\d+%3a(.+)$/) { - rename $_, "$1_$2"; - } -} -' - -echo "Removing old binary packages." -# (Versions of packages which have more then 5 versions and those old versions are more than one month old.) -perl -e ' -for (<*.deb>) { - m/(.*?)_/; - $fs{$1}++; -} -while (($f,$n) = each(%fs)) { - next if $n <= 5; - unlink for grep {-M > 30} <$f_*.deb>; -} -' - -echo "Making packages list." -cd /Volumes/fink/ -perl -MFink::Scanpackages -e "Fink::Scanpackages->scan('dists/10.$MAC_VERSION/main/binary-darwin-$ARCH/');" | gzip - > dists/10.$MAC_VERSION/main/binary-darwin-$ARCH/Packages.gz - -echo "Copying to repository all info files." -rm -f /Volumes/fink/dists/10.$MAC_VERSION/main/finkinfo/* -cp $FINK_ROOT/fink/dists/biolab/main/finkinfo/* /Volumes/fink/dists/10.$MAC_VERSION/main/finkinfo/ - -echo "Making an archive of all info files." -cd $FINK_ROOT/fink/dists/biolab/main/finkinfo/ -tar -czf /Volumes/fink/dists/10.$MAC_VERSION/main/finkinfo/all.tgz * - -echo "Setting permissions." -chmod -R +r /Volumes/fink/dists/10.$MAC_VERSION/main/finkinfo/ - -echo "Removing unnecessary source archives." -perl -e " -for () { - m/_(.+)-\\d+_darwin-$ARCH\\.deb/; - \$versions{\$1} = 1; -} -for () { - m/.+-(.+)\\.tgz/; - next if \$versions{\$1} or -M() < 30; - unlink; -} -" - Index: install-scripts/mac/update-all-scripts.sh =================================================================== --- install-scripts/mac/update-all-scripts.sh (revision 10505) +++ install-scripts/mac/update-all-scripts.sh (revision 10342) @@ -22,7 +22,4 @@ curl --silent --output bundle-inject-pypi.sh https://bitbucket.org/biolab/orange/raw/tip/install-scripts/mac/bundle-inject-pypi.sh curl --silent --output dailyrun-bundleonly-hg.sh https://bitbucket.org/biolab/orange/raw/tip/install-scripts/mac/dailyrun-bundleonly-hg.sh -curl --silent --output fink-daily-build-packages.sh https://bitbucket.org/biolab/orange/raw/tip/install-scripts/mac/fink-daily-build-packages.sh -curl --silent --output fink-register-info.sh https://bitbucket.org/biolab/orange/raw/tip/install-scripts/mac/fink-register-info.sh -curl --silent --output build-source.sh https://bitbucket.org/biolab/orange/raw/tip/install-scripts/mac/build-source.sh chmod +x *.sh Index: setup.py =================================================================== --- setup.py (revision 10507) +++ setup.py (revision 10471) @@ -676,4 +676,6 @@ version = full_version """ + global VERSION + FULLVERSION = VERSION if os.path.exists('.hg'): @@ -698,4 +700,7 @@ finally: a.close() + + if not ISRELEASED: + VERSION = FULLVERSION def setup_package(): Index: source/orange/lib_learner.cpp =================================================================== --- source/orange/lib_learner.cpp (revision 10517) +++ source/orange/lib_learner.cpp (revision 10377) @@ -2099,10 +2099,9 @@ bool setPrefixRules; bool optimizeBetasFlag; - float penalty = 0.01f; - - if (!PyArg_ParseTuple(args, "O&fffO&|O&iiO&O&:RuleClassifier.call", cc_RuleList, &rules, &minSignificance, &minBeta, &penalty, pt_ExampleGenerator, &gen, pt_weightByGen(gen), &weightID, &setPrefixRules, &optimizeBetasFlag, cc_Classifier, &classifier, cc_DistributionList, &probList)) - return PYNULL; - - TRuleClassifier *rc = new TRuleClassifier_logit(rules, minSignificance, minBeta, penalty, gen, weightID, classifier, probList, setPrefixRules, optimizeBetasFlag); + + if (!PyArg_ParseTuple(args, "O&ffO&|O&iiO&O&:RuleClassifier.call", cc_RuleList, &rules, &minSignificance, &minBeta, pt_ExampleGenerator, &gen, pt_weightByGen(gen), &weightID, &setPrefixRules, &optimizeBetasFlag, cc_Classifier, &classifier, cc_DistributionList, &probList)) + return PYNULL; + + TRuleClassifier *rc = new TRuleClassifier_logit(rules, minSignificance, minBeta, gen, weightID, classifier, probList,setPrefixRules, optimizeBetasFlag); PRuleClassifier ruleClassifier = rc; // ruleClassifier = new SELF_AS(TRuleClassifier)(rules, gen, weightID); Index: source/orange/rulelearner.cpp =================================================================== --- source/orange/rulelearner.cpp (revision 10531) +++ source/orange/rulelearner.cpp (revision 8735) @@ -45,5 +45,5 @@ : weightID(0), quality(ILLEGAL_FLOAT), - complexity(0), + complexity(-1), coveredExamples(NULL), coveredExamplesLength(-1), @@ -93,5 +93,5 @@ TRule::~TRule() -{ delete coveredExamples; } +{ delete coveredExamples; } bool TRule::operator ()(const TExample &ex) @@ -986,6 +986,6 @@ improved/rule->classDistribution->atint(targetClass) > min_improved_perc*0.01 && quality > (aprioriProb + 1e-3)) -// futureQuality = quality; - futureQuality = 1.0 + quality; + futureQuality = quality; +// futureQuality = 1.0 + quality; else { PDistribution oldRuleDist = rule->classDistribution; @@ -1016,5 +1016,5 @@ futureQuality = -1.0; else { - futureQuality = 0.0; + futureQuality = 0.0; PEITERATE(ei, rule->examples) { if ((*ei).getClass().intV != targetClass) @@ -1023,22 +1023,11 @@ continue; } - /* float x = ((*ei)[probVar].floatV-quality); //*rule->classDistribution->abs; - if ((*ei)[probVar].floatV > quality) + float x = ((*ei)[probVar].floatV-quality); //*rule->classDistribution->abs; + if ((*ei)[probVar].floatV > quality) x *= (1.0-quality)/(bestQuality-quality); x /= sqrt(quality*(1.0-quality)); // rule->classDistribution->abs* - futureQuality += log(1.0-max(1e-12,1.0-2*zprob(x))); */ - float x; - if ((*ei)[probVar].floatV > quality) - { - x = 1.0 - ((*ei)[probVar].floatV-quality) / (bestQuality-quality); - } - else - { - x = 1.0 + quality - (*ei)[probVar].floatV; - } - futureQuality += x; - } -// futureQuality = 1.0 - exp(futureQuality); - futureQuality /= rule->classDistribution->atint(targetClass); + futureQuality += log(1.0-max(1e-12,1.0-2*zprob(x))); + } + futureQuality = 1.0 - exp(futureQuality); } } @@ -1970,17 +1959,17 @@ { PITERATE(TIntList, ind, ruleIndices[rule_i]) - { - float bestQuality = 0.0; - PITERATE(TIntList, fr, prefixRules) { - if (rules->at(*fr)->call(examples->at(*ind)) && rules->at(*fr)->quality > bestQuality) { - bestQuality = rules->at(*fr)->quality; - p[getClassIndex(rules->at(*fr))][*ind] = rules->at(*fr)->quality; - for (int ci=0; cidomain->classVar->noOfValues(); ci++) - if (ci!=getClassIndex(rules->at(*fr))) - p[ci][*ind] = (1.0-rules->at(*fr)->quality)/(examples->domain->classVar->noOfValues()-1); - break; - } - } - } + { + float bestQuality = 0.0; + PITERATE(TIntList, fr, prefixRules) { + if (rules->at(*fr)->call(examples->at(*ind)) && rules->at(*fr)->quality > bestQuality) { + bestQuality = rules->at(*fr)->quality; + p[getClassIndex(rules->at(*fr))][*ind] = rules->at(*fr)->quality; + for (int ci=0; cidomain->classVar->noOfValues(); ci++) + if (ci!=getClassIndex(rules->at(*fr))) + p[ci][*ind] = (1.0-rules->at(*fr)->quality)/(examples->domain->classVar->noOfValues()-1); + break; + } + } + } } @@ -2029,10 +2018,10 @@ void TLogitClassifierState::setPrefixRule(int rule_i) //, int position) { - prefixRules->push_back(rule_i); - setFixed(rule_i); - updateFixedPs(rule_i); - betas[rule_i] = 0.0; - computeAvgProbs(); - computePriorProbs(); + prefixRules->push_back(rule_i); + setFixed(rule_i); + updateFixedPs(rule_i); + betas[rule_i] = 0.0; + computeAvgProbs(); + computePriorProbs(); } @@ -2041,5 +2030,5 @@ {} -TRuleClassifier_logit::TRuleClassifier_logit(PRuleList arules, const float &minSignificance, const float &minBeta, const float &penalty, PExampleTable anexamples, const int &aweightID, const PClassifier &classifier, const PDistributionList &probList, bool setPrefixRules, bool optimizeBetasFlag) +TRuleClassifier_logit::TRuleClassifier_logit(PRuleList arules, const float &minSignificance, const float &minBeta, PExampleTable anexamples, const int &aweightID, const PClassifier &classifier, const PDistributionList &probList, bool setPrefixRules, bool optimizeBetasFlag) : TRuleClassifier(arules, anexamples, aweightID), minSignificance(minSignificance), @@ -2047,9 +2036,9 @@ setPrefixRules(setPrefixRules), optimizeBetasFlag(optimizeBetasFlag), - minBeta(minBeta), - penalty(penalty) + minBeta(minBeta) { initialize(probList); float step = 2.0; + minStep = (float)0.01; // initialize prior betas @@ -2062,10 +2051,10 @@ if (setPrefixRules) { - bool changed = setBestPrefixRule(); - while (changed) { - if (optimizeBetasFlag) - optimizeBetas(); - changed = setBestPrefixRule(); - } + bool changed = setBestPrefixRule(); + while (changed) { + if (optimizeBetasFlag) + optimizeBetas(); + changed = setBestPrefixRule(); + } } @@ -2218,8 +2207,8 @@ wsig = sig; PITERATE(TRuleList, ri, rules) { - float maxDiff = (*ri)->classDistribution->atint(getClassIndex(*ri))/(*ri)->classDistribution->abs; - maxDiff -= (*ri)->quality; - wsig->push_back(maxDiff); - float n = (*ri)->examples->numberOfExamples(); + float maxDiff = (*ri)->classDistribution->atint(getClassIndex(*ri))/(*ri)->classDistribution->abs; + maxDiff -= (*ri)->quality; + wsig->push_back(maxDiff); + float n = (*ri)->examples->numberOfExamples(); float a = n*(*ri)->quality; float b = n*(1.0-(*ri)->quality); @@ -2283,8 +2272,8 @@ if ((*ri)->call(*ei)) { //int vv = (*ei).getClass().intV; - //if ((*ei).getClass().intV == getClassIndex(*ri)) - coverages[getClassIndex(*ri)][j] += 1.0; + //if ((*ei).getClass().intV == getClassIndex(*ri)) + coverages[getClassIndex(*ri)][j] += 1.0; } - j++; + j++; } i++; @@ -2324,15 +2313,13 @@ float step = 0.1f; + float gamma = 0.01f; float error = 1e+20f; float old_error = 1e+21f; - int nsteps = 0; while (old_error > error || step > 0.00001) { // reduce step if improvement failed - nsteps++; - if (old_error < error && nsteps > 20 || nsteps > 1000) + if (old_error < error) { - nsteps = 0; step /= 10; finalState->copyTo(currentState); @@ -2348,14 +2335,7 @@ float der = 0.0; - if (currentState->avgProb->at(i) > rules->at(i)->quality) - { - der += pow(rules->at(i)->quality - currentState->avgProb->at(i),2); - der -= 2 * (rules->at(i)->quality - currentState->avgProb->at(i)) * currentState->betas[i]; - } - else - der -= 2 * (rules->at(i)->quality - currentState->avgProb->at(i)); - der += 2 * penalty * currentState->betas[i]; -// if (currentState->avgProb->at(i) > rules->at(i)->quality) -// der = max(0.01f / step, der); + if (currentState->avgProb->at(i) < rules->at(i)->quality) + der -= rules->at(i)->quality - currentState->avgProb->at(i); + der += 2*gamma*currentState->betas[i]; currentState->newBeta(i,max(0.0f, currentState->betas[i]-step*der)); } @@ -2363,10 +2343,7 @@ error = 0; for (int i=0; isize(); i++) { -// if (currentState->avgProb->at(i) < rules->at(i)->quality) - if (currentState->avgProb->at(i) > rules->at(i)->quality) - error += pow(rules->at(i)->quality - currentState->avgProb->at(i),2) * currentState->betas[i]; - else - error += pow(rules->at(i)->quality - currentState->avgProb->at(i),2); - error += penalty*pow(currentState->betas[i],2); + if (currentState->avgProb->at(i) < rules->at(i)->quality) + error += rules->at(i)->quality - currentState->avgProb->at(i); + error += gamma*pow(currentState->betas[i],2); } //printf("error = %4.4f\n", error); @@ -2376,7 +2353,111 @@ //printf("\n"); } - finalState->copyTo(currentState); } + +/* +void TRuleClassifier_logit::updateRuleBetas2(float step_) +{ + + stabilizeAndEvaluate(step_,-1); + PLogitClassifierState finalState, tempState; + currentState->copyTo(finalState); + + float step = 2.0; + int changed; + float worst_underestimate, underestimate; + float auc = currentState->getAUC(); + float brier = currentState->getBrierScore(); + float temp_auc, temp_brier; + int worst_rule_index; + while (step > 0.001) + { + step /= 2; + changed = 0; + while (changed < 100) + { + changed = 0; + worst_underestimate = (float)0.01; + worst_rule_index = -1; + // find rule with greatest underestimate in probability + for (int i=0; isize(); i++) { + if (currentState->avgProb->at(i) >= rules->at(i)->quality) + continue; + if (skipRule[i]) + continue; + + underestimate = (rules->at(i)->quality - currentState->avgProb->at(i));//*rules->at(i)->classDistribution->abs; + // if under estimate error is big enough + if (underestimate > worst_underestimate) + { + worst_underestimate = underestimate; + worst_rule_index = i; + } + } + if (worst_rule_index > -1) + { + currentState->newBeta(worst_rule_index,currentState->betas[worst_rule_index]+step); + if (currentState->avgProb->at(worst_rule_index) > rules->at(worst_rule_index)->quality) + { + finalState->copyTo(currentState); + changed = 100; + } + else + { + stabilizeAndEvaluate(step,-1); + temp_auc = currentState->getAUC(); + temp_brier = currentState->getBrierScore(); + if (temp_auc >= auc && temp_brier < brier) + { + currentState->copyTo(finalState); + changed = 0; + auc = temp_auc; + brier = temp_brier; + } + else + changed ++; + } + // } + } + else + { + changed = 100; + finalState->copyTo(currentState); + } + } + } + finalState->copyTo(currentState); +} +*/ + +void TRuleClassifier_logit::stabilizeAndEvaluate(float & step, int last_changed_rule_index) +{ + PLogitClassifierState tempState; + currentState->copyTo(tempState); + bool changed = true; + while (changed) + { + changed = false; + for (int i=0; isize(); i++) + { + if (currentState->avgProb->at(i) > (rules->at(i)->quality + 0.01) && currentState->betas[i] > 0.0 && + i != last_changed_rule_index) + { + float new_beta = currentState->betas[i]-step > 0 ? currentState->betas[i]-step : 0.0; + currentState->newBeta(i,new_beta); + if (currentState->avgProb->at(i) < rules->at(i)->quality + 1e-6) + { + tempState->copyTo(currentState); + } + else + { + currentState->copyTo(tempState); + changed = true; + } + } + } + } +} + void TRuleClassifier_logit::addPriorClassifier(const TExample &ex, double * priorFs) { @@ -2435,14 +2516,14 @@ bool foundPrefixRule = false; float bestQuality = 0.0; - PITERATE(TRuleList, rs, prefixRules) { - if ((*rs)->call(ex) && (*rs)->quality > bestQuality) { - bestQuality = (*rs)->quality; - dist->setint(getClassIndex(*rs),(*rs)->quality); - for (int ci=0; cidomain->classVar->noOfValues(); ci++) - if (ci!=getClassIndex(*rs)) - dist->setint(ci,(1.0-(*rs)->quality)/(examples->domain->classVar->noOfValues()-1)); - foundPrefixRule = true; - break; - } + PITERATE(TRuleList, rs, prefixRules) { + if ((*rs)->call(ex) && (*rs)->quality > bestQuality) { + bestQuality = (*rs)->quality; + dist->setint(getClassIndex(*rs),(*rs)->quality); + for (int ci=0; cidomain->classVar->noOfValues(); ci++) + if (ci!=getClassIndex(*rs)) + dist->setint(ci,(1.0-(*rs)->quality)/(examples->domain->classVar->noOfValues()-1)); + foundPrefixRule = true; + break; + } } if (foundPrefixRule) @@ -2464,7 +2545,7 @@ if ((*r)->call(cexample)) { if (getClassIndex(*r) == i) - f += (*b); + f += (*b); else if (getClassIndex(*r) == res->noOfElements()-1) - f -= (*b); + f -= (*b); } dist->addint(i,exp(f)); Index: source/orange/rulelearner.hpp =================================================================== --- source/orange/rulelearner.hpp (revision 10517) +++ source/orange/rulelearner.hpp (revision 8735) @@ -482,14 +482,12 @@ __REGISTER_CLASS - PRuleList rules; //P - PExampleTable examples; //P + PRuleList rules; + PExampleTable examples; int weightID; float eval, **f, **p, *betas, *priorBetas; bool *isExampleFixed; - PFloatList avgProb; //P - PFloatList avgPriorProb; //P - PIntList *ruleIndices; //P - PIntList prefixRules; //P + PFloatList avgProb, avgPriorProb; + PIntList *ruleIndices, prefixRules; TLogitClassifierState(PRuleList, PExampleTable, const int &); @@ -517,22 +515,18 @@ PDomain domain; //P Domain PFloatList ruleBetas; //P Rule betas + float minStep; //P minimal step value float minSignificance; //P minimum requested significance for betas. float minBeta; //P minimum beta by rule to be included in the model. - bool setPrefixRules; //P should we order prefix rules ? - bool optimizeBetasFlag; //P should we assign betas to rules ? - float penalty; //P + bool setPrefixRules; // P should we order prefix rules ? + bool optimizeBetasFlag; // P should we assign betas to rules ? PClassifier priorClassifier; //P prior classifier used if provided - PLogitClassifierState currentState; //P + PLogitClassifierState currentState; bool *skipRule; - // standard deviations of rule quality - PFloatList wsd; //P - PFloatList wavgCov; //P - PFloatList wSatQ; //P - PFloatList wsig; //P + PFloatList wsd, wavgCov, wSatQ, wsig; // standard deviations of rule quality PRuleList prefixRules; //P rules that trigger before logit sum. TRuleClassifier_logit(); - TRuleClassifier_logit(PRuleList rules, const float &minSignificance, const float &minBeta, const float &penalty, PExampleTable examples, const int &weightID = 0, const PClassifier &classifer = NULL, const PDistributionList &probList = NULL, bool setPrefixRules = false, bool optimizeBetasFlag = true); + TRuleClassifier_logit(PRuleList rules, const float &minSignificance, const float &minBeta, PExampleTable examples, const int &weightID = 0, const PClassifier &classifer = NULL, const PDistributionList &probList = NULL, bool setPrefixRules = false, bool optimizeBetasFlag = true); void initialize(const PDistributionList &); Index: source/orangeqt/CMakeLists.txt =================================================================== --- source/orangeqt/CMakeLists.txt (revision 10515) +++ source/orangeqt/CMakeLists.txt (revision 9252) @@ -9,5 +9,5 @@ find_package(SIP REQUIRED) find_package(PythonLibrary REQUIRED) -#find_package(OpenGL) +find_package(OpenGL) find_package(Numpy) include(SIPMacros) @@ -29,7 +29,7 @@ multicurve.cpp plot.cpp -# plot3d.cpp -# canvas3d.cpp -# glextensions.cpp + plot3d.cpp + canvas3d.cpp + glextensions.cpp ) qt4_automoc(${orangeqt_SRCS}) @@ -51,6 +51,5 @@ set(SIP_EXTRA_OPTIONS "-k") -#set(SIP_EXTRA_LINK_LIBRARIES ${QT_QTCORE_LIBRARY} ${QT_QTGUI_LIBRARY} ${QT_QTOPENGL_LIBRARY} ${OPENGL_gl_LIBRARY}) -set(SIP_EXTRA_LINK_LIBRARIES ${QT_QTCORE_LIBRARY} ${QT_QTGUI_LIBRARY}) +set(SIP_EXTRA_LINK_LIBRARIES ${QT_QTCORE_LIBRARY} ${QT_QTGUI_LIBRARY} ${QT_QTOPENGL_LIBRARY} ${OPENGL_gl_LIBRARY}) set(SIP_EXTRA_SOURCES ${orangeqt_SRCS}) Index: source/orangeqt/orangeqt.sip =================================================================== --- source/orangeqt/orangeqt.sip (revision 10515) +++ source/orangeqt/orangeqt.sip (revision 9057) @@ -2,10 +2,10 @@ %Import QtGui/QtGuimod.sip -/*%Import QtOpenGL/QtOpenGLmod.sip*/ +%Import QtOpenGL/QtOpenGLmod.sip %Include types.sip %Include plot.sip -/*%Include plot3d.sip -%Include canvas3d.sip*/ +%Include plot3d.sip +%Include canvas3d.sip %Include plotitem.sip %Include point.sip