Why do i have NAN values in the confusion matrix only in the validation test?

I wanted to create neural network for binary classification for dataset with input matrix with size [9 981] and output matrix [1 981]and this is the code that i used

   rng('default');
inputs = patientInputs;
targets = patientTargets;
x = mapminmax(inputs); 
t=targets; 
 trainFcn = 'trainbr'; 
% Create a Pattern Recognition Network
hiddenLayerSize =10;
net = patternnet(hiddenLayerSize,trainFcn);

 net.divideFcn = 'dividerand';  % Divide data randomly
 net.divideMode = 'sample';  % Divide up every sample
net.divideParam.trainRatio = 70/100;
net.divideParam.valRatio = 15/100;
net.divideParam.testRatio = 15/100;
net.performFcn = 'mse';  

 % Choose Plot Functions
 % For a list of all plot functions type: help nnplot
 net.plotFcns = {'plotperform','plottrainstate','ploterrhist', ...
    'plotconfusion', 'plotroc'};

% Train the Network
net= configure(net,x,t);
[net,tr] = train(net,x,t);

 y = net(x);
 e = gsubtract(t,y);
 performance = perform(net,t,y)
tind = vec2ind(t);
yind = vec2ind(y);
percentErrors = sum(tind ~= yind)/numel(tind);

 % Recalculate Training, Validation and Test Performance
 trainTargets = t .* tr.trainMask{1};
valTargets = t .* tr.valMask{1};
testTargets = t .* tr.testMask{1};
trainPerformance = perform(net,trainTargets,y)
valPerformance = perform(net,valTargets,y)
testPerformance = perform(net,testTargets,y)

% View the Network
view(net)

At first i used the default trainFcn 'trainscg' then i tried to use 'trainbr' the accuracy improved but i got NAN values in the confusion matrix only in the validation test as you can see it here

Can anyone help me please?

Answer 1

[Greg Heath]

MATLAB doesn't allow a validation set for trainbr because they think it isn't necessary for generalization.

Although they are correct, I have found in the literature that using BOTH trainbr and a validation set is better in most cases.

Although you cannot force trainbr to have validation set, I think you can work around it using trainlm and/or trainscg with regularization.

Hope this helps

Thank you for formally accepting my answer

Greg

PS: I don't remember details.

[afef]

When i was reading matlab documentation about trainbr (link = https://www.mathworks.com/help/nnet/ref/trainbr.html) i found that validation is enabled by setting max_fail to 6 and i tried this and it give me this in the cofusion matrix. I don't know am i right ?

[muhammed shames]

Hello sir can you explain me how you got rid of that NaN values ? Ive been working on it but it didnt work by just setting the max fail to 6 !

[Ismail Cem Tüzün]

when I set max fail to 6 even 10, it worked. I also re-assigned my epoch number. May be it also affected the system, not sure. Thanks for the information, btw