|
Daniel@0
|
1 <html>
|
|
Daniel@0
|
2 <head>
|
|
Daniel@0
|
3 <title>
|
|
Daniel@0
|
4 Netlab Reference Manual gtminit
|
|
Daniel@0
|
5 </title>
|
|
Daniel@0
|
6 </head>
|
|
Daniel@0
|
7 <body>
|
|
Daniel@0
|
8 <H1> gtminit
|
|
Daniel@0
|
9 </H1>
|
|
Daniel@0
|
10 <h2>
|
|
Daniel@0
|
11 Purpose
|
|
Daniel@0
|
12 </h2>
|
|
Daniel@0
|
13 Initialise the weights and latent sample in a GTM.
|
|
Daniel@0
|
14
|
|
Daniel@0
|
15 <p><h2>
|
|
Daniel@0
|
16 Synopsis
|
|
Daniel@0
|
17 </h2>
|
|
Daniel@0
|
18 <PRE>
|
|
Daniel@0
|
19 net = gtminit(net, options, data, samptype)
|
|
Daniel@0
|
20 net = gtminit(net, options, data, samptype, lsampsize, rbfsampsize)
|
|
Daniel@0
|
21 </PRE>
|
|
Daniel@0
|
22
|
|
Daniel@0
|
23
|
|
Daniel@0
|
24 <p><h2>
|
|
Daniel@0
|
25 Description
|
|
Daniel@0
|
26 </h2>
|
|
Daniel@0
|
27 <CODE>net = gtminit(net, options, data, samptype)</CODE> takes a GTM <CODE>net</CODE>
|
|
Daniel@0
|
28 and generates a sample of latent data points and sets the centres (and
|
|
Daniel@0
|
29 widths if appropriate) of
|
|
Daniel@0
|
30 <CODE>net.rbfnet</CODE>.
|
|
Daniel@0
|
31
|
|
Daniel@0
|
32 <p>If the <CODE>samptype</CODE> is <CODE>'regular'</CODE>, then regular grids of latent
|
|
Daniel@0
|
33 data points and RBF centres are created. The dimension of the latent data
|
|
Daniel@0
|
34 space must be
|
|
Daniel@0
|
35 1 or 2. For one-dimensional latent space, the <CODE>lsampsize</CODE> parameter
|
|
Daniel@0
|
36 gives the number of latent points and the <CODE>rbfsampsize</CODE> parameter
|
|
Daniel@0
|
37 gives the number of RBF centres. For a two-dimensional latent space,
|
|
Daniel@0
|
38 these parameters must be vectors of length 2 with the number of points
|
|
Daniel@0
|
39 in each of the x and y directions to create a rectangular grid. The
|
|
Daniel@0
|
40 widths of the RBF basis functions are set by a call to <CODE>rbfsetfw</CODE>
|
|
Daniel@0
|
41 passing <CODE>options(7)</CODE> as the scaling parameter.
|
|
Daniel@0
|
42
|
|
Daniel@0
|
43 <p>If the <CODE>samptype</CODE> is <CODE>'uniform'</CODE> or <CODE>'gaussian'</CODE> then the
|
|
Daniel@0
|
44 latent data is found by sampling from a uniform or
|
|
Daniel@0
|
45 Gaussian distribution correspondingly. The RBF basis function parameters
|
|
Daniel@0
|
46 are set
|
|
Daniel@0
|
47 by a call to <CODE>rbfsetbf</CODE> with the <CODE>data</CODE> parameter
|
|
Daniel@0
|
48 as dataset and the <CODE>options</CODE> vector.
|
|
Daniel@0
|
49
|
|
Daniel@0
|
50 <p>Finally, the output layer weights of the RBF are initialised by
|
|
Daniel@0
|
51 mapping the mean of the latent variable to the mean of the target variable,
|
|
Daniel@0
|
52 and the L-dimensional latent variale variance to the variance of the
|
|
Daniel@0
|
53 targets along the first L principal components.
|
|
Daniel@0
|
54
|
|
Daniel@0
|
55 <p><h2>
|
|
Daniel@0
|
56 See Also
|
|
Daniel@0
|
57 </h2>
|
|
Daniel@0
|
58 <CODE><a href="gtm.htm">gtm</a></CODE>, <CODE><a href="gtmem.htm">gtmem</a></CODE>, <CODE><a href="pca.htm">pca</a></CODE>, <CODE><a href="rbfsetbf.htm">rbfsetbf</a></CODE>, <CODE><a href="rbfsetfw.htm">rbfsetfw</a></CODE><hr>
|
|
Daniel@0
|
59 <b>Pages:</b>
|
|
Daniel@0
|
60 <a href="index.htm">Index</a>
|
|
Daniel@0
|
61 <hr>
|
|
Daniel@0
|
62 <p>Copyright (c) Ian T Nabney (1996-9)
|
|
Daniel@0
|
63
|
|
Daniel@0
|
64
|
|
Daniel@0
|
65 </body>
|
|
Daniel@0
|
66 </html> |
