comparison toolboxes/FullBNT-1.0.7/nethelp3.3/fevbayes.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 fevbayes
5 </title>
6 </head>
7 <body>
8 <H1> fevbayes
9 </H1>
10 <h2>
11 Purpose
12 </h2>
13 Evaluate Bayesian regularisation for network forward propagation.
14
15 <p><h2>
16 Synopsis
17 </h2>
18 <PRE>
19 extra = fevbayes(net, y, a, x, t, x_test)
20 [extra, invhess] = fevbayes(net, y, a, x, t, x_test, invhess)
21 </PRE>
22
23
24 <p><h2>
25 Description
26 </h2>
27 <CODE>extra = fevbayes(net, y, a, x, t, x_test)</CODE> takes a network data structure
28 <CODE>net</CODE> together with a set of hidden unit activations <CODE>a</CODE> from
29 test inputs <CODE>x_test</CODE>, training data inputs <CODE>x</CODE> and <CODE>t</CODE> and
30 outputs a matrix of extra information <CODE>extra</CODE> that consists of
31 error bars (variance)
32 for a regression problem or moderated outputs for a classification problem.
33 The optional argument (and return value)
34 <CODE>invhess</CODE> is the inverse of the network Hessian
35 computed on the training data inputs and targets. Passing it in avoids
36 recomputing it, which can be a significant saving for large training sets.
37
38 <p>This is called by network-specific functions such as <CODE>mlpevfwd</CODE> which
39 are needed since the return values (predictions and hidden unit activations)
40 for different network types are in different orders (for good reasons).
41
42 <p><h2>
43 See Also
44 </h2>
45 <CODE><a href="mlpevfwd.htm">mlpevfwd</a></CODE>, <CODE><a href="rbfevfwd.htm">rbfevfwd</a></CODE>, <CODE><a href="glmevfwd.htm">glmevfwd</a></CODE><hr>
46 <b>Pages:</b>
47 <a href="index.htm">Index</a>
48 <hr>
49 <p>Copyright (c) Ian T Nabney (1996-9)
50
51
52 </body>
53 </html>