Brief explanation of all the training function in nntool

There are multiple training function to choose from on the neural network tool (trainlm, trainscg, traingdx, etc) and I was wondering about each of them; what scenarios are best to use on each and on what kind of datasets?

Also, if I have 122 elements inputs, each of them having value of either, '1' or '0' or '-1' and the target has 1 element - either '1' or '0', which function should I use?

Thanks.

Answer 1

[Greg Heath]

If you have a regression or curve-fitting problem use training function FITNET. Accept all defaults and start with the code from the documentation examples in

 help fitnet

and

 doc fitnet

If you have a classification or pattern-recognition problem start with the sample code in

 help patternnet

and

 doc patternnet

For more complete examples you can search in both the NEWSREADER in comp.soft-sys.matlab and ANSWERS.

Start the searches with the keywords

 greg fitnet

or

 greg patternnet

Hope this helps.

Thank you for formally accepting my answer

Greg