How to arrange input for CNN ?

Hi, I'm trying to use convolutional neural network for load forecasting with 26 feature to forecast 1 output (response), I have prepared the rack of input and output as

cnn_inp = zeros(featureDimension,1,numResponses,length(xtr(:,1)));
cnn_op = zeros(numResponses,1,numResponses,length(xtr(:,1)));
cat_cnn_op=categorical(cnn_op);

because the output need to be categorical, then I standardize the output between [0 1] and give some bound in which the result is classified into ['max' 'normal' 'min']

I have arranged the output as categorical vector but it still can't work.

Error using trainNetwork (line 154)
Invalid training data. Y must be a vector of categorical responses.

Error in method_comparison (line 370)
    net = trainNetwork(cnn_inp,cat_cnn_op,layers,options);

Caused by:
    Error using nnet.internal.cnn.util.NetworkDataValidator/assertCategoricalResponseVector (line 136)
    Invalid training data. Y must be a vector of categorical responses.

if anyone can help me to resolve this issue, I'll much appriciate. Thank you

Answer 1

[Sai Bhargav Avula]

Hi,

Categorical is the last step you do before passing it to the trainNetwork function. Try creating the categorical array of the output just before your trainNetwork not during the intilization.