|
Daniel@0
|
1 <html>
|
|
Daniel@0
|
2 <head>
|
|
Daniel@0
|
3 <title>
|
|
Daniel@0
|
4 Netlab Reference Manual rbfprior
|
|
Daniel@0
|
5 </title>
|
|
Daniel@0
|
6 </head>
|
|
Daniel@0
|
7 <body>
|
|
Daniel@0
|
8 <H1> rbfprior
|
|
Daniel@0
|
9 </H1>
|
|
Daniel@0
|
10 <h2>
|
|
Daniel@0
|
11 Purpose
|
|
Daniel@0
|
12 </h2>
|
|
Daniel@0
|
13 Create Gaussian prior and output layer mask for RBF.
|
|
Daniel@0
|
14
|
|
Daniel@0
|
15 <p><h2>
|
|
Daniel@0
|
16 Synopsis
|
|
Daniel@0
|
17 </h2>
|
|
Daniel@0
|
18 <PRE>
|
|
Daniel@0
|
19 [mask, prior] = rbfprior(rbfunc, nin, nhidden, nout, aw2, ab2)</PRE>
|
|
Daniel@0
|
20
|
|
Daniel@0
|
21
|
|
Daniel@0
|
22 <p><h2>
|
|
Daniel@0
|
23 Description
|
|
Daniel@0
|
24 </h2>
|
|
Daniel@0
|
25 <CODE>[mask, prior] = rbfprior(rbfunc, nin, nhidden, nout, aw2, ab2)</CODE>
|
|
Daniel@0
|
26 generates a vector
|
|
Daniel@0
|
27 <CODE>mask</CODE> that selects only the output
|
|
Daniel@0
|
28 layer weights. This is because most uses of RBF networks in a Bayesian
|
|
Daniel@0
|
29 context have fixed basis functions with the output layer as the only
|
|
Daniel@0
|
30 adjustable parameters. In particular, the Neuroscale output error function
|
|
Daniel@0
|
31 is designed to work only with this mask.
|
|
Daniel@0
|
32
|
|
Daniel@0
|
33 <p>The return value
|
|
Daniel@0
|
34 <CODE>prior</CODE> is a data structure,
|
|
Daniel@0
|
35 with fields <CODE>prior.alpha</CODE> and <CODE>prior.index</CODE>, which
|
|
Daniel@0
|
36 specifies a Gaussian prior distribution for the network weights in an
|
|
Daniel@0
|
37 RBF network. The parameters <CODE>aw2</CODE> and <CODE>ab2</CODE> are all
|
|
Daniel@0
|
38 scalars and represent the regularization coefficients for two groups
|
|
Daniel@0
|
39 of parameters in the network corresponding to
|
|
Daniel@0
|
40 second-layer weights, and second-layer biases
|
|
Daniel@0
|
41 respectively. Then <CODE>prior.alpha</CODE> represents a column vector of
|
|
Daniel@0
|
42 length 2 containing the parameters, and <CODE>prior.index</CODE> is a matrix
|
|
Daniel@0
|
43 specifying which weights belong in each group. Each column has one
|
|
Daniel@0
|
44 element for each weight in the matrix, using the standard ordering as
|
|
Daniel@0
|
45 defined in <CODE>rbfpak</CODE>, and each element is 1 or 0 according to
|
|
Daniel@0
|
46 whether the weight is a member of the corresponding group or not.
|
|
Daniel@0
|
47
|
|
Daniel@0
|
48 <p><h2>
|
|
Daniel@0
|
49 See Also
|
|
Daniel@0
|
50 </h2>
|
|
Daniel@0
|
51 <CODE><a href="rbf.htm">rbf</a></CODE>, <CODE><a href="rbferr.htm">rbferr</a></CODE>, <CODE><a href="rbfgrad.htm">rbfgrad</a></CODE>, <CODE><a href="evidence.htm">evidence</a></CODE><hr>
|
|
Daniel@0
|
52 <b>Pages:</b>
|
|
Daniel@0
|
53 <a href="index.htm">Index</a>
|
|
Daniel@0
|
54 <hr>
|
|
Daniel@0
|
55 <p>Copyright (c) Ian T Nabney (1996-9)
|
|
Daniel@0
|
56
|
|
Daniel@0
|
57
|
|
Daniel@0
|
58 </body>
|
|
Daniel@0
|
59 </html> |
