Error in running one-dimensional CNN regression model

Hi,

I'm new to training a CNN regression model, and I need to build one to predict temperature time series data using climate proxy data.

I'm not sure what are functions of various layers, so I'm building a simple CNN model first.

For example, I've temperature time series of length 100 (Y_train=100x1 double) and 45 climate proxy data (X_train=45x100 double), so the input feature number is 45.

I've set my CNN model architecture as follows:

function y_pred = cnn(X_train, X_val, Y_train)
    layers = [sequenceInputLayer(45,'MinLength',100)    
        convolution1dLayer(2,10)    
        reluLayer
        maxPooling1dLayer(2,'Stride',2)
        reluLayer
        fullyConnectedLayer(1)
        reluLayer
        regressionLayer];
    
    options = trainingOptions('sgdm', ...
        MaxEpochs=500, ...
        Verbose=false, ...
        Plots='none');

    net = trainNetwork(X_train, Y_train, layers, options);
    y_pred = predict(net, X_val);
end

When I ran the code, it produces this error message:

I'm unsure where I've done wrong.

Answer 1

[Prasannavenkatesh]

Hi Marvin,

To solve the error of incompatible input and output sequence lengths, you can try replacing the first layer, i.e, the sequence input layer to sequenceInputLayer(45,'MinLength',1). Hope this solves your problem.