source: orange/docs/widgets/rst/regression/pade.rst @ 11778:ecd4beec2099

Revision 11778:ecd4beec2099, 2.9 KB checked in by Ales Erjavec <ales.erjavec@…>, 5 months ago (diff)

Use new SVG icons in the widget documentation.

Line 
1.. _Pade:
2
3Pade
4====
5
6.. image:: ../../../../Orange/OrangeWidgets/Regression/icons/Pade.svg
7
8Replaces a continuous class with a derivative or a MQC by one or more
9continuous attributes.
10
11Signals
12-------
13
14Inputs:
15   - Classified Examples (ExampleTableWithClass)
16      Input data set.
17
18
19Outputs:
20   - Classified Examples (ExampleTableWithClass)
21      Output data set.
22
23
24Description
25-----------
26
27This widget implements several techniques for assessing partial derivatives
28of the class variable for the given set of examples. The derivative is appended
29to the example table as a new class attribute. The widget can compute either
30quantitative derivative by a chosen continuous attribute or a qualitative
31derivative by one or more attributes.
32
33The widget is implemented to cache some data. After, for instance, computing
34the derivatives by ``x`` and ``y`` separately, the widget has already
35stored all the data to produce the derivatives by both in a moment.
36
37.. image:: images/Pade.png
38
39The :obj:`Attributes` box lists all continuous attributes and lets the user
40select the attribute by which she wants to compute the qualitative derivative.
41The selection is important only when the widget actually outputs a qualitative
42derivative (this depends on other settings, described below). Buttons
43:obj:`All` and :obj:`None` select the entire list and nothing.
44
45Derivatives by more than one attribute are mathematically questionable, and
46computing by many attributes can be slow and messy. Methods that are based on
47triangulation will include all attributes in the triangulation, regardless of
48the selection, but then compute only the selected derivatives.
49
50Box :obj:`Method` determines the used method and its settings. Available
51methods are :obj:`First triangle`, :obj:`Star Regression`,
52:obj:`Univariate Star Regression` and :obj:`Tube Regression`. First triangle is
53unsuitable for data with non-negligible noise. Star regression seems to perform
54rather poor; the quantitative derivatives it computes are even theoretically
55wrong. Univariate Star Regression will handle noise well, but also work well
56for very complex functions (like sin(x)sin(y) across several periods). Tube
57regression is very noise resistant, which can lead it to oversimplify the
58model, yet it is the only method that does not use the triangulation and is
59thus capable of handling discrete attributes, unknown values and large number
60of dimensions. It may be slow when the number of examples is very large.
61Detailed description of these methods can be found in Zabkar and Demsar's
62papers.
63
64:obj:`Ignore differences below` lets the user set a threshold for qualitative
65derivatives.
66
67The widget can also put some data in meta attributes: the
68:obj:`Qualitative constraint`, as described above,
69:obj:`Derivatives of selected attributes` and the
70:obj:`Original class attribute`.
71
72The changes take effect and the widget start processing when :obj:`Apply`
73is hit.
Note: See TracBrowser for help on using the repository browser.