Validation Loss = Nan

Question

aryan ramesh 2022-2-6

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/1644325-validation-loss-nan

评论： aryan ramesh 2022-2-8

采纳的回答： yanqi liu

在 MATLAB Online 中打开

Hello, I'm attempting to utilize lstm to categorize data but the validation loss Is Nan.

I reduced the learning rates to 1e-12 but I am still receiving Nan results.

Appreciate any guidance.

Best Regards,

options = trainingOptions("sgdm", ...
    "MaxEpochs",400, ...
    "InitialLearnRate",0.000000000001, ...
    "Shuffle", 'never', ...
    "Plots","training-progress",...
    "ValidationData",{XValidation,YValidation},...
    'ValidationFrequency',1);
%%
layers = [ ...
    sequenceInputLayer(1)
    bilstmLayer(100,"OutputMode","last")
    fullyConnectedLayer(2)
    softmaxLayer
    classificationLayer];
%  displaySequence(tones_cell{1}, label1{1})
net =  trainNetwork(XTrain,labelTrain, layers, options )
YPred = classify(net,XTest);