Hi Wissal,
When predicting with a NARX network using new external inputs, it's important to prepare the input data so that the network’s internal states are properly updated. Instead of calling:
ypred = netc(xnew, xic, aic)
or
ypred = netc(xnew)
directly which may lead to errors or uniform outputs, you should use the ‘preparets’ function to format your new input data and initialize the network's delay states. For example:
% Prepare the new external input (xnew) along with empty target sequences
[xnewPrepared, xic, aic] = preparets(netc, xnew, {});
% Predict using the prepared inputs and initial conditions
ypred = netc(xnewPrepared, xic, aic);
This approach ensures that the ‘NARX’ network is provided with correctly formatted data and properly initialized states, avoiding errors and preventing the output from being constant across columns.
For a better understanding of the above solution, refer to the following MATLAB documentation:
