how to include four hidden layers by taking away LSTM.

I am having a code that implements LSTM layer as below:

inputSize = 12;

numHiddenUnits1 = 48;

numHiddenUnits2 = 48;

numHiddenUnits3 = 48;

numHiddenUnits4 = 48;

numClasses = 12;

layers = [ ...

sequenceInputLayer(inputSize)

lstmLayer(numHiddenUnits1,'OutputMode','sequence')

lstmLayer(numHiddenUnits2,'OutputMode','sequence')

lstmLayer(numHiddenUnits3,'OutputMode','sequence')

lstmLayer(numHiddenUnits4,'OutputMode','sequence')

fullyConnectedLayer(numClasses)

reluLayer

regressionLayer]

Now I want to implement four hidden layers without LSTM. So anyone can help me how to modify it.

Answer 1

[Aparajith Raghuvir]

I understand you would like to add 4 fully connected hidden layers without LSTM.

The same fullyConnectedLayer() function can be used with your choice of hyperparemeters to achieve your aim. Please refer to this for the documentation of the abovementioned function.

I hope this helps.

[jaah navi]

Yes. I want four hidden layers with (48 hidden nodes each) without LSTM.

I tried with the following lines

inputSize = 12;

numHiddenUnits1 = 48;

numHiddenUnits2 = 48;

numHiddenUnits3 = 48;

numHiddenUnits4 = 48;

numClasses = 12

layers = [ ...

sequenceInputLayer(inputSize)

fullyConnectedLayer(numHiddenUnits1,numHiddenUnits2,numHiddenUnits3,numHiddenUnits4,numClasses)

reluLayer

regressionLayer] but I am getting error. Could you please help me to overcome the error.

[Aparajith Raghuvir]

For 4 hidden layers, you can try the following:

inputSize = 12;
numHiddenUnits1 = 48;
numHiddenUnits2 = 48;
numHiddenUnits3 = 48;
numHiddenUnits4 = 48;
numClasses = 12
 layers = [ ...
sequenceInputLayer(inputSize)
fullyConnectedLayer(numHiddenUnits1)
reluLayer
fullyConnectedLayer(numHiddenUnits2)
reluLayer
fullyConnectedLayer(numHiddenUnits3)
reluLayer
fullyConnectedLayer(numHiddenUnits4)
reluLayer
fullyConnectedLayer(numClasses)
reluLayer
regressionLayer]

Please refer to this link for in depth documentation on all the deep learning layers and also some examples that you can view on how to use them.

I hope this helps.

[jaah navi]

Thanks. It works.

[jaah navi]

Could you please help me how to add sine, cosine, tanh activation by replacing reLu layer for the above code.