I have sequence data in a cell with 64x6512 for my CNN although it says input size mismatch on my the first convolutional layer, how do I fix this error?

I am trying to build a CNN network that will take in sequence data. And I don't know how to fix the input size mismatch.

XTrain is a cell array of 450x1 containing 64x6152 Matrices

While YTrain is a categorical array that pairs with XTrain.

The CNN Network:

XTrain=ClassComp;

YTrain=Label;

layers = [

sequenceInputLayer(64,"Name","sequence")

convolution2dLayer([50 1],10,"Name","conv_1","Padding","same")

convolution2dLayer([3 1],10,"Name","conv_2","Padding","same")

maxPooling2dLayer([5 5],"Name","maxpool","Padding","same")

fullyConnectedLayer(64,"Name","fc_1")

dropoutLayer(0.5,"Name","dropout")

fullyConnectedLayer(10,"Name","fc_2")

softmaxLayer("Name","softmax")

classificationLayer("Name","classoutput")];

options = trainingOptions('sgdm', ...

'ExecutionEnvironment','cpu', ...

'GradientThreshold',1, ...

'MaxEpochs',10, ...

'SequenceLength','longest', ...

'Shuffle','never', ...

'Verbose',0, ...

'Plots','training-progress');

net = trainNetwork(ClassComp,Label,layers,options);

The Error

Caused by:

Layer 'conv_1': Input size mismatch. Size of input to this layer is different from the expected input size.

Inputs to this layer:

from layer 'sequence' (size 64(C) × 1(B) × 1(T))

Answer 1

[Anshika Chaurasia]

Hi Javier,

The solution is to apply sequenceFoldingLayer and sequenceUnfoldingLayer.

Refer to this example to better understand where to apply these layers.

Hope it helps!

[Javier Castillo]

Hi Anshika,

It seems that the input size mismatch error was fixed although this error would pop up which I am uncertain of how to fix with the responses of my training data. My response array is a 450x1 categorical array which corresponds to my 450x1 training data.

Response Data:

The error that it would show was:

Error using trainNetwork (line 184)

Invalid training data. Responses must be a cell array of categorical response sequences.

Which then I used this type of response instead:

Response Data:

Which gave me the same error.