Convolutional 1d net

Hi everyone! I am trying to reproduce the convolutional network described in http://lemt.ufrj.br/pdf/pedro.pdf (Table 4.1). I am using as input a matrix 370x2401 and as a target another matrix 370x2401. The network's architecture is:

When I try to train the network appears the next message: Error using trainNetwork>iAssertCategoricalResponseVector (line 598) Y must be a vector of categorical responses.

I have used reshape function applied to input: X=reshape(input, 1,2401,1,370); When trying I used: trainNetwork(X,Y,layers,opts); where Y is the target matrix 370x2401. What am I doing wrong?

Thanks beforehand!

Answer 1

[Sanjana Ramakrishnan]

Please note that for image and sequence-to-label classification problems, Y must be a categorical vector of labels. Refer the below link: https://www.mathworks.com/help/nnet/ref/trainnetwork.html?searchHighlight=trainNetwork&s_tid=doc_srchtitle#inputarg_Y

In your code, the function 'trainNetwork' is expecting "Y" to be a categorical array, not a matrix. Consider using "categorical" function to convert "Y" into categorical array as below:

>> Y = categorical(Y);

Refer the below link:

https://www.mathworks.com/help/matlab/ref/categorical.html

[Ash Ash]

Thank you!