I have a problem in the trainNetwork for Xtrain and Ytrain it gives me X and Y must have the same number of observations.

XTrain = importdata('C:\Users\m7mod\Documents\MATLAB\TrainM.mat'); % size of XTrain = 9 * 44 YTrain = categorical([1 1 1 1 0 0 0 0 0]'); % size of YTrain = 9 * 1 layers = [ ... imageInputLayer([44 1]) convolution2dLayer(5,20) reluLayer fullyConnectedLayer(10) softmaxLayer classificationLayer()] options = trainingOptions('sgdm'); XNew = zeros(size(XTrain,1),1,1,size(XTrain,2)); XNew(:,1,1,:) = XTrain(:,:); net = trainNetwork(XNew,YTrain',layers,options);

[Greg Heath]

Please check your dictionary for the word

                      FORMAT

Answer 1

[Ameer Hamza]

You are facing the problem because you are trying to use imageInputLayer and convolution2dLayer which will only work if your input sample have at least 2 non-singleton dimensions (i.e. m*n and m*n*k will work but 1*m or m*1 will not work). For a single dimension data (as in your case 1*44), you can use sequenceInputLayers. For your case, if you can change the layers combination as shown in following script snippet the code will work

XTrain = importdata('C:\Users\m7mod\Documents\MATLAB\TrainM.mat'); % size of XTrain = 9 * 44 
YTrain = categorical([1 1 1 1 0 0 0 0 0]'); % size of YTrain = 9 * 1 
layers = [ ...
    sequenceInputLayer(44)
    fullyConnectedLayer(2)
    softmaxLayer
    classificationLayer];
options = trainingOptions('sgdm'); 
net = trainNetwork(XTrain',YTrain',layers,options);

[mahmoud Bassiouni]

Thanks for the answer I just have one more question sequenceInputLayer command works only in matlab R2018a

[Ameer Hamza]

The documentation page state that it was introduced in 2017b.

If you are using earlier version then you can use fitnet and train as

XTrain = importdata('C:\Users\m7mod\Documents\MATLAB\TrainM.mat'); % size of XTrain = 9 * 44 
YTrain = [1 1 1 1 0 0 0 0 0]'; % size of YTrain = 9 * 1 
net = fitnet([10 10]);
net = train(net, XTrain',YTrain');

[mahmoud Bassiouni]

XTrain = AllTrainCel(1:200000,:)'; 4 * 200000

YTrain = categorical([1 0 -1 -2]'); % 4 * 1;

layers = [ ...

sequenceInputLayer(200000)

%reluLayer

LSTMLayer

fullyConnectedLayer(4)

softmaxLayer

classificationLayer];

options = trainingOptions('sgdm');

net = trainNetwork(XTrain',YTrain',layers,options);

net = fitnet([10 10]);

XTest = AllTestCel(1:200000,:)';

YTest = categorical([1 0 -1 -2]');

[YPred] = classify(net,XTest);

The problem is in the classification stage the classify doesnt work it. It needs more parameters although it work in the examples with two parameters only

Answer 2

[mahmoud Bassiouni]

The problem is in the classification stage the classify doesnt work it. It needs more parameters although it work in the examples of deep learning with two parameters only

XTrain = AllTrainCel(1:200000,:)'; 4 * 200000

YTrain = categorical([1 0 -1 -2]'); % 4 * 1;

layers = [ ...

sequenceInputLayer(200000)

%reluLayer

LSTMLayer

fullyConnectedLayer(4)

softmaxLayer

classificationLayer];

options = trainingOptions('sgdm');

net = trainNetwork(XTrain',YTrain',layers,options);

net = fitnet([10 10]);

XTest = AllTestCel(1:200000,:)';

YTest = categorical([1 0 -1 -2]');

[YPred] = classify(net,XTest);