annotate toolboxes/FullBNT-1.0.7/nethelp3.3/mlpevfwd.htm @ 0:cc4b1211e677 tip

initial commit to HG from Changeset: 646 (e263d8a21543) added further path and more save "camirversion.m"
author Daniel Wolff
date Fri, 19 Aug 2016 13:07:06 +0200
parents
children
rev   line source
Daniel@0 1 <html>
Daniel@0 2 <head>
Daniel@0 3 <title>
Daniel@0 4 Netlab Reference Manual mlpevfwd
Daniel@0 5 </title>
Daniel@0 6 </head>
Daniel@0 7 <body>
Daniel@0 8 <H1> mlpevfwd
Daniel@0 9 </H1>
Daniel@0 10 <h2>
Daniel@0 11 Purpose
Daniel@0 12 </h2>
Daniel@0 13 Forward propagation with evidence for MLP
Daniel@0 14
Daniel@0 15 <p><h2>
Daniel@0 16 Synopsis
Daniel@0 17 </h2>
Daniel@0 18 <PRE>
Daniel@0 19
Daniel@0 20 [y, extra] = mlpevfwd(net, x, t, x_test)
Daniel@0 21 [y, extra, invhess] = mlpevfwd(net, x, t, x_test, invhess)
Daniel@0 22 </PRE>
Daniel@0 23
Daniel@0 24
Daniel@0 25 <p><h2>
Daniel@0 26 Description
Daniel@0 27 </h2>
Daniel@0 28 <CODE>y = mlpevfwd(net, x, t, x_test)</CODE> takes a network data structure
Daniel@0 29 <CODE>net</CODE> together with the input <CODE>x</CODE> and target <CODE>t</CODE> training data
Daniel@0 30 and input test data <CODE>x_test</CODE>.
Daniel@0 31 It returns the normal forward propagation through the network <CODE>y</CODE>
Daniel@0 32 together with a matrix <CODE>extra</CODE> which consists of error bars (variance)
Daniel@0 33 for a regression problem or moderated outputs for a classification problem.
Daniel@0 34 The optional argument (and return value)
Daniel@0 35 <CODE>invhess</CODE> is the inverse of the network Hessian
Daniel@0 36 computed on the training data inputs and targets. Passing it in avoids
Daniel@0 37 recomputing it, which can be a significant saving for large training sets.
Daniel@0 38
Daniel@0 39 <p><h2>
Daniel@0 40 See Also
Daniel@0 41 </h2>
Daniel@0 42 <CODE><a href="fevbayes.htm">fevbayes</a></CODE><hr>
Daniel@0 43 <b>Pages:</b>
Daniel@0 44 <a href="index.htm">Index</a>
Daniel@0 45 <hr>
Daniel@0 46 <p>Copyright (c) Ian T Nabney (1996-9)
Daniel@0 47
Daniel@0 48
Daniel@0 49 </body>
Daniel@0 50 </html>