Your target set must have one label for each sample, so it must contains 2300 elements. Here is an easy way to build one. Start by numbering your classes from 1 to 23. Make sure that your X Array is sorted by class, meaning all the samples from the first class, then all the samples from the second class, and so on. Then:
% Classes from 1 to 23
y = 1:23;
% 100 samples per class
y = repmat(y, 100, 1);
% Reshape to obtain a vector
y = reshape(y, 1, numel(y));
At this stage, you should have a 1-by-2300 vector with numeric class labels from 1 to 23. This is what you want if you use the Statistics and Machine Learning tToolbox.
Now, I suppose you intend to use the Neural Network Toolbox. In that case the syntax is a bit different: the class labels are in a 23-by-2300 matrix, each column being a sample, with the row corresponding to the class getting value '1' and the rest '0'. To build it from the previous syntax, type:
y = full(ind2vec(y));
Good luck!
