Invalid training data in LSTM

I have given the following dimensions data for sequence to label classification using LSTM.....

xtrain = 56724 x 1 cell (each cell is having 1 x 2560 double)

ytrain = 56724 x 1 categorical

I am getting the following error:

Invalid training data. Predictors must be a N-by-1 cell array of sequences, where N is the number of

sequences. All sequences must have the same feature dimension and at least one time step.

Can anyone solve my issue.

Answer 1

[Harsha Priya Daggubati]

Hi,

The proper input format to train a LSTM Network should be as follows,

XTrain = N x 1 cell

YTrain = N x 1 , where N denotes the Number of Sequences/Training Examples.

Each cell in Training set can have variable number of columns, but a definite set of rows. It implies each cell is of the form K x L, where K is the number of input features and L is the input data for that feature.

Here in your case, I doubt there is an issue with your XTrain cell. Try modifying your XTrain to match the input format specified above.

Hope this helps!

[swetha annangi]

Thank you Harsha. I verified my XTrain cells....there is a mistake in few cells. i rectified it.