Neural Network Toolbox

plotbr

Purpose

Plot network performance for Bayesian regularization training.

Syntax

plotbr(TR,name,epoch)

Description

plotbr(tr,name,epoch) takes these inputs,

• TR -- Training record returned by train

  name -- Training function name, default = ''

  epoch -- Number of epochs, default = length of training record

and plots the training sum squared error, the sum squared weight, and the effective number of parameters.

Examples

Here are input values P and associated targets T.

• p = [-1:.05:1];
  t = sin(2*pi*p)+0.1*randn(size(p));
  

The code below creates a network and trains it on this problem.

• net=newff([-1 1],[20,1],{'tansig','purelin'},'trainbr');
  [net,tr] = train(net,p,t);
  

During training plotbr was called to display the training record. You can also call plotbr directly with the final training record TR, as shown below.

• plotbr(tr)
  

normr

plotep

© 1994-2005 The MathWorks, Inc.