Adam Optimizer with feedforward nueral networks

Hello, is there any way to use Adam optimizer to train a neural network with the "train" fucntion? Or a way to use this implementation ( https://www.mathworks.com/matlabcentral/fileexchange/61616-adam-stochastic-gradient-descent-optimization ) to train the network?

Thanks in advance.

[Abdelwahab Afifi]

Have you get the answer ?

[Bob]

did anyone of you guys got the answer?

Answer 1

[Hrishikesh Borate]

Hi,

It’s my understanding that you want to use Adam optimizer to train a neural network. This can be done using trainNetwork function, and setting the appropriate Training Options.

For eg.,

[XTrain,~,YTrain] = digitTrain4DArrayData;
layers = [ ...
    imageInputLayer([28 28 1])
    convolution2dLayer(12,25)
    reluLayer
    fullyConnectedLayer(1)
    regressionLayer];
options = trainingOptions('adam', ...
    'InitialLearnRate',0.001, ...
    'GradientThreshold',1, ...
    'Verbose',false, ...
    'Plots','training-progress');
net = trainNetwork(XTrain,YTrain,layers,options);
[XTest,~,YTest] = digitTest4DArrayData;
YPred = predict(net,XTest);
rmse = sqrt(mean((YTest - YPred).^2))

For more information, refer to trainNetwork.

[Abdelwahab Afifi]

'trainNetwork' is used for Deep learning neural network. But I think he wanna use 'Adam optimizer' to train shallow neural network using 'train' function.