Mercurial > hg > camir-aes2014
comparison toolboxes/FullBNT-1.0.7/nethelp3.3/rbfprior.htm @ 0:e9a9cd732c1e tip
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 rbfprior | |
| 5 </title> | |
| 6 </head> | |
| 7 <body> | |
| 8 <H1> rbfprior | |
| 9 </H1> | |
| 10 <h2> | |
| 11 Purpose | |
| 12 </h2> | |
| 13 Create Gaussian prior and output layer mask for RBF. | |
| 14 | |
| 15 <p><h2> | |
| 16 Synopsis | |
| 17 </h2> | |
| 18 <PRE> | |
| 19 [mask, prior] = rbfprior(rbfunc, nin, nhidden, nout, aw2, ab2)</PRE> | |
| 20 | |
| 21 | |
| 22 <p><h2> | |
| 23 Description | |
| 24 </h2> | |
| 25 <CODE>[mask, prior] = rbfprior(rbfunc, nin, nhidden, nout, aw2, ab2)</CODE> | |
| 26 generates a vector | |
| 27 <CODE>mask</CODE> that selects only the output | |
| 28 layer weights. This is because most uses of RBF networks in a Bayesian | |
| 29 context have fixed basis functions with the output layer as the only | |
| 30 adjustable parameters. In particular, the Neuroscale output error function | |
| 31 is designed to work only with this mask. | |
| 32 | |
| 33 <p>The return value | |
| 34 <CODE>prior</CODE> is a data structure, | |
| 35 with fields <CODE>prior.alpha</CODE> and <CODE>prior.index</CODE>, which | |
| 36 specifies a Gaussian prior distribution for the network weights in an | |
| 37 RBF network. The parameters <CODE>aw2</CODE> and <CODE>ab2</CODE> are all | |
| 38 scalars and represent the regularization coefficients for two groups | |
| 39 of parameters in the network corresponding to | |
| 40 second-layer weights, and second-layer biases | |
| 41 respectively. Then <CODE>prior.alpha</CODE> represents a column vector of | |
| 42 length 2 containing the parameters, and <CODE>prior.index</CODE> is a matrix | |
| 43 specifying which weights belong in each group. Each column has one | |
| 44 element for each weight in the matrix, using the standard ordering as | |
| 45 defined in <CODE>rbfpak</CODE>, and each element is 1 or 0 according to | |
| 46 whether the weight is a member of the corresponding group or not. | |
| 47 | |
| 48 <p><h2> | |
| 49 See Also | |
| 50 </h2> | |
| 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> | |
| 52 <b>Pages:</b> | |
| 53 <a href="index.htm">Index</a> | |
| 54 <hr> | |
| 55 <p>Copyright (c) Ian T Nabney (1996-9) | |
| 56 | |
| 57 | |
| 58 </body> | |
| 59 </html> |