annotate toolboxes/FullBNT-1.0.7/nethelp3.3/demard.htm @ 0:e9a9cd732c1e tip

first hg version after svn
author wolffd
date Tue, 10 Feb 2015 15:05:51 +0000
parents
children
rev   line source
wolffd@0 1 <html>
wolffd@0 2 <head>
wolffd@0 3 <title>
wolffd@0 4 Netlab Reference Manual demard
wolffd@0 5 </title>
wolffd@0 6 </head>
wolffd@0 7 <body>
wolffd@0 8 <H1> demard
wolffd@0 9 </H1>
wolffd@0 10 <h2>
wolffd@0 11 Purpose
wolffd@0 12 </h2>
wolffd@0 13 Automatic relevance determination using the MLP.
wolffd@0 14
wolffd@0 15 <p><h2>
wolffd@0 16 Synopsis
wolffd@0 17 </h2>
wolffd@0 18 <PRE>
wolffd@0 19 demmlp1</PRE>
wolffd@0 20
wolffd@0 21
wolffd@0 22 <p><h2>
wolffd@0 23 Description
wolffd@0 24 </h2>
wolffd@0 25 This script demonstrates the technique of automatic relevance
wolffd@0 26 determination (ARD) using a synthetic problem having three input
wolffd@0 27 variables: <CODE>x1</CODE> is sampled uniformly from the range (0,1) and has
wolffd@0 28 a low level of added Gaussian noise, <CODE>x2</CODE> is a copy of <CODE>x1</CODE>
wolffd@0 29 with a higher level of added noise, and <CODE>x3</CODE> is sampled randomly
wolffd@0 30 from a Gaussian distribution. The single target variable is determined
wolffd@0 31 by <CODE>sin(2*pi*x1)</CODE> with additive Gaussian noise. Thus <CODE>x1</CODE> is
wolffd@0 32 very relevant for determining the target value, <CODE>x2</CODE> is of some
wolffd@0 33 relevance, while <CODE>x3</CODE> is irrelevant. The prior over weights is
wolffd@0 34 given by the ARD Gaussian prior with a separate hyper-parameter for
wolffd@0 35 the group of weights associated with each input. A multi-layer
wolffd@0 36 perceptron is trained on this data, with re-estimation of the
wolffd@0 37 hyper-parameters using <CODE>evidence</CODE>. The final values for the
wolffd@0 38 hyper-parameters reflect the relative importance of the three inputs.
wolffd@0 39
wolffd@0 40 <p><h2>
wolffd@0 41 See Also
wolffd@0 42 </h2>
wolffd@0 43 <CODE><a href="demmlp1.htm">demmlp1</a></CODE>, <CODE><a href="demev1.htm">demev1</a></CODE>, <CODE><a href="mlp.htm">mlp</a></CODE>, <CODE><a href="evidence.htm">evidence</a></CODE><hr>
wolffd@0 44 <b>Pages:</b>
wolffd@0 45 <a href="index.htm">Index</a>
wolffd@0 46 <hr>
wolffd@0 47 <p>Copyright (c) Ian T Nabney (1996-9)
wolffd@0 48
wolffd@0 49
wolffd@0 50 </body>
wolffd@0 51 </html>