I created a
LVQ network with 41 neurons (I don't know I did right or not, I have 125000 samples with 41 features), in fact I'm using
NSL kdd dataset. here is my code in matlab, I set it train for 1 epoch, but though its performance shows convergence but when I test it by this line:
y = net(norm_x); , it label all the test samples the same! it means it doesnt recognize at all! am I doing something wrong? or may I test wrongly?is there any other wat to test my trained network?
another thing that I do is normalizing the dataset.
%%%calculating average feature instanc:
% for i=1:41
%%%calculating standard deviation feature instance:
standard_deviation(j) = sum(power((x(:,j)- mean_features(j)),2))/41;
%%%calculating normalized instance:
net.trainParam.epochs = 1;
net = lvqnet(41); %creating lvq network with 41 hidden neuron
net = train(net,norm_x,target); %training the network
y = net(norm_x);
perf = perform(net,y,target);