Feedforward neural networks 3. FFNet versus discriminant classifier

You may want to compare the FFNet classifier with a discriminant classifier. Unlike the FFNet, a discriminant classifier does not need any iterative procedure in the learning phase and can be used immediately after creation for classification. The following three simple steps will give you the confusion matrix based on discriminant analysis:

1. Select the PatternList and the Categories together and choose To Discriminant. A newly created Discriminant will appear.
2. Select the Discriminant and the PatternList together and choose To Categories.... A newly created Categories will appear.
3. Select the two appropriate Categories and choose To Confusion. A newly created Confusion will appear. After pushing the Info button, the info window will show you the fraction correct.

See also the Discriminant analysis tutorial for more information.

Links to this page

© djmw, April 26, 2004