Multiple input feedforwardnet neural network toolbox

To whom it may concern;

Although I have spent time on the following problem, I could NOT solved it.

I am trying to approximate a function in the following structure by using feedforwardnet:

f(x1,x2)=y

where 'x1' and 'x2' are the inputs while 'y' is the output.

For this I have 'x1', 'x2' and 'y' as 1000 element vectors.

It is possible to create a net by using the following code;

newff([-1 1; -1,1],[10 1]) this returns the following net.

However since newff is obsolete I am looking for a way to do the same thing by feedforwardnet.

So could you please show me how to generate the same net by using feedforwardnet?

Yours Sincerely.

Answer 1

[Greg Heath]

Specialized functions NEWFIT (regression and curve-fitting) and NEWPR (classification and pattern-recognition) automatically call the generic function NEWFF. All have been obsolete since 2010 but are still available.

The current specialized functions FITNET (regression and curve-fitting) and PATTERNNET (classification and pattern-recognition) automatically call the generic function FEEDFORWARDNET.

Since you have a regression problem, use FITNET. The documentation and a corresponding simple example can be obtained from the documentation commands

 help fitnet

 doc fitnet

More practical examples can be obtained by searching the NEWSGROUP and ANSWERS using

 greg fitnet

If searching for a good choice of number of hidden nodes

 greg fitnet Hmin:dH:Hmax

and/or if searching for a good choice of random initial weights and data division

 greg fitnet Ntrials

Hope this helps.

Thank you for formally accepting my answer

Greg