source: orange/orange/doc/widgets/Regression/Earth.htm @ 9563:46bb78f9f167

Revision 9563:46bb78f9f167, 2.5 KB checked in by ales_erjavec <ales.erjavec@…>, 2 years ago (diff)

Added basic documentation for Earth widget.

Line 
1<html>
2<head>
3<title>Earth</title>
4<link rel=stylesheet href="../../../style.css" type="text/css" media=screen>
5<link rel=stylesheet href="../../../style-print.css" type="text/css" media=print></link>
6</head>
7
8<body>
9
10<h1>Earth Learner</h1>
11
12<img class="screenshot" src="../icons/Unknown.png">
13<p>Earth Learner</p>
14
15<h2>Channels</h2>
16
17<h3>Inputs</h3>
18
19<dl class=attributes>
20<dt>Data (Table)</dt>
21<dd>A table with training instances</dd>
22</dl>
23
24<h3>Outputs</h3>
25<dl class=attributes>
26<dt>Learner</dt>
27<dd>The Earth learning algorithm with parameters as specified in the dialog.</dd>
28
29<dt>Predictor</dt>
30<dd>Trained regressor (a subtype of Classifier)</dd>
31</dl>
32
33<p>Signal <code>Predictor</code> sends data only if the input signal
34<code>Data</code> is present.</p>
35
36<h2>Description</h2>
37
38<p>This widget constructs a Earth learning algorithm (an implementation of the
39<a href="http://en.wikipedia.org/wiki/Multivariate_adaptive_regression_splines">
40MARS - Multivariate Adaptive Regression Splines</a>). As all widgets for
41classification and regression, this widget provides a learner and
42classifier/regressor on the output. Learner is a learning algorithm with
43settings as specified by the user. It can be fed into widgets for testing
44learners, for instance <code>Test Learners</code>.</p>
45
46<img class="leftscreenshot" src="Earth.png" alt="Earth Widget" border=0/>
47
48<p>Learner can be given a name under which it will appear in, say, <code>Test Learners</code>.
49The default name is "Earth Learner".</p>
50
51<p>The <code>Max. term degree</code> parameter specifies the degree of the terms induced in
52the forward pass. For instance, if set to <code>1</code> the resulting model will contain
53only linear terms.</p>
54
55<p>The <code>Max. terms</code> specifies how many terms can be induces in the forward pass.
56A special value <code>Automatic</code> instructs the learner to set the limit automatically
57based on the dimensionality of the data
58(<code>min(200, max(20, 2 * NumberOfAttributes)) + 1</code>).</p> 
59
60<p>The <code>Knot penalty</code> is used in the pruning pass (hinge function penalty for
61the GCV calculation)</p>
62
63<p>After changing one or more settings, you need to push <span class="option">Apply</span>,
64which will put the new learner on the output and, if the training examples are given,
65construct a new predictor and output it as well.</p>
66
67<h2>Examples</h2>
68
69<p>Lets use the learner to train a model on a data subset and test it on unseen instances.</p>
70
71<img class="screenshot" src="Earth-Schema.png">
72
73</body>
74</html>
Note: See TracBrowser for help on using the repository browser.