How can I plot the plotperf() with training validation and testing data set on command line

Hi I have trained a MLP neural network with 1000 epoch, using the Levenberg-Marquardt as training algorithm with msereg as the performance function I have divided the data as 40% 40% 20% (training validation testing).

How can I plot the plotperf() with training validation and testing data set on command line. Regards

Answer 1

[Greg Heath]

It depends on what version you are using.

help plotperf and doc plotperf yield

Obsoleted in R2008b NNET 6.0. Last used in R2007b NNET 5.1.

The example in help and doc use

net = newff(minmax(P),[4 1],{'tansig','tansig'});

whereas, both help newff and doc newff yield

Obsoleted in R2010b NNET 7.0. Last used in R2010a NNET 6.0.4.

However, those refer to a later obsolete version of newff that has the syntax

net = newff(P,T,S,TF,BTF,BLF,PF,IPF,OPF,DDF);

and different defaults.

The latest version of the NNTBX uses fitnet for regression/curvefitting, (patternnet for classification/pattern=recognition) and plotperform.

Since there are at least three different code scenarios with different defaults, you have to specify your version and include relevant code.

Hope this helps.

Greg