It would depend on the particular 'operation'. If your 'operation' works on a 10*50 matrix, and for example produces a scalar value, then you could record this for all 10 data sets:
size_of_dataset = 10;
n_datasets = size(myMatrix, 1) / size_of_dataset; % check how many observations are there
res = zeros(1, n_datasets); % initialise results as zeros
for k = 1:n_datasets
res(k) = myOperation( myMatrix(k-1+(1:size_of_dataset),:) ); % get all data for the dataset
end
It would be easier to keep track of the separate datasets if it were stored in a 10*10*50 array instead, then you could loop through a 'layer' each time.
