How to save a trained Neural Network with NNTools and classify a new set of data based on it? I don't want to use the testing/validation percentage in the GUI

Hi,

I have used NNstart to train a NN based on my data set. I have successfully done that and can see the confusion matrices too. Now I want to save this trained NN and input a new data set for testing the classification result. I don't want to use the GUI's tool to assign a percentage of my training data for testing and validation. Can you experts shed some light on how to achieve that? 1. how to save the trained NN? 2. how to get the confusion matrix based on a new testing set that is analysed against the saved NN? 3. What are the internal parameters/functions Matlab uses for the trained NN?

Cheers, M.

[Greg Heath]

I deleted your question

"Is this even possible?"

because you placed it in the answer box.

Greg

Answer 1

[Mehrdad Azh]

write this function in Matlab:

%%%%%%%%%%%%%%%%%%%%%%%%%%

function iris_datasets()

[x,t] = iris_dataset;%this is a data set from Matlab

net = patternnet(5);%hidden layers

net = train(net,x,t);%trains the NN

save net;

end

%%%%%%%%%%%%%%%%%%%%%%%%%%

then in command prompt test a new data set like "new" on the same trained NN:

new2=[5.3;3.3;1.2;0.4];% this is similar to the first column of x, which is classified under the first class in t

sim(net,new2);%simulates NN based on your test set

ans =

    0.9996
    0.0004
    0.0000