Mercurial > hg > camir-aes2014

comparison toolboxes/FullBNT-1.0.7/nethelp3.3/somtrain.htm @ 0:e9a9cd732c1e tip

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
first hg version after svn
author wolffd
date Tue, 10 Feb 2015 15:05:51 +0000
parents
children
comparison
equal deleted inserted replaced
-1:000000000000 0:e9a9cd732c1e
1 <html>
2 <head>
3 <title>
4 Netlab Reference Manual somtrain
5 </title>
6 </head>
7 <body>
8 <H1> somtrain
9 </H1>
10 <h2>
11 Purpose
12 </h2>
13 Kohonen training algorithm for SOM.
14
15 <p><h2>
16 Synopsis
17 </h2>
18 <PRE>
19
20 net = somtrain{net, options, x)
21 </PRE>
22
23
24 <p><h2>
25 Description
26 </h2>
27 <CODE>net = somtrain{net, options, x)</CODE> uses Kohonen's algorithm to
28 train a SOM. Both on-line and batch algorithms are implemented.
29 The learning rate (for on-line) and neighbourhood size decay linearly.
30 There is no error function minimised during training (so there is
31 no termination criterion other than the number of epochs), but the
32 sum-of-squares is computed and returned in <CODE>options(8)</CODE>.
33
34 <p>The optional parameters have the following interpretations.
35
36 <p><CODE>options(1)</CODE> is set to 1 to display error values; also logs learning
37 rate <CODE>alpha</CODE> and neighbourhood size <CODE>nsize</CODE>.
38 Otherwise nothing is displayed.
39
40 <p><CODE>options(5)</CODE> determines whether the patterns are sampled randomly
41 with replacement. If it is 0 (the default), then patterns are sampled
42 in order. This is only relevant to the on-line algorithm.
43
44 <p><CODE>options(6)</CODE> determines if the on-line or batch algorithm is
45 used. If it is 1
46 then the batch algorithm is used. If it is 0
47 (the default) then the on-line algorithm is used.
48
49 <p><CODE>options(14)</CODE> is the maximum number of iterations (passes through
50 the complete pattern set); default 100.
51
52 <p><CODE>options(15)</CODE> is the final neighbourhood size; default value is the
53 same as the initial neighbourhood size.
54
55 <p><CODE>options(16)</CODE> is the final learning rate; default value is the same
56 as the initial learning rate.
57
58 <p><CODE>options(17)</CODE> is the initial neighbourhood size; default 0.5*maximum
59 map size.
60
61 <p><CODE>options(18)</CODE> is the initial learning rate; default 0.9. This parameter
62 must be positive.
63
64 <p><h2>
65 Examples
66 </h2>
67 The following example performs on-line training on a SOM in two stages:
68 ordering and convergence.
69 <PRE>
70
71 net = som(nin, [8, 7]);
72 options = foptions;
73
74 <p>% Ordering phase
75 options(1) = 1;
76 options(14) = 50;
77 options(18) = 0.9; % Initial learning rate
78 options(16) = 0.05; % Final learning rate
79 options(17) = 8; % Initial neighbourhood size
80 options(15) = 1; % Final neighbourhood size
81 net2 = somtrain(net, options, x);
82
83 <p>% Convergence phase
84 options(14) = 400;
85 options(18) = 0.05;
86 options(16) = 0.01;
87 options(17) = 0;
88 options(15) = 0;
89 net3 = somtrain(net2, options, x);
90 </PRE>
91
92
93 <p><h2>
94 See Also
95 </h2>
96 <CODE><a href="kmeans.htm">kmeans</a></CODE>, <CODE><a href="som.htm">som</a></CODE>, <CODE><a href="somfwd.htm">somfwd</a></CODE><hr>
97 <b>Pages:</b>
98 <a href="index.htm">Index</a>
99 <hr>
100 <p>Copyright (c) Ian T Nabney (1996-9)
101
102
103 </body>
104 </html>