cell_rows = arrayfun(@(ROW) horzcat(YourCell{ROW,:}), 1:size(YourCell,1), 'uniform', 0);
all_values = horzcat(cell_rows{:});
[G, uvals] = findgroups(all_values);
num_vals = length(uvals);
row_lens = cellfun(@length, cell_rows);
G_by_row = mat2cell(G, 1, row_lens);
counts = cellfun(@(R) [uvals(:), accumarray(R(:), 1, [num_vals 1])], G_by_row, 'uniform', 0);
The result will be a cell array with 63 entries. Each entry will be an N x 2 table, where N is the number of unique values over the entire matrix (not the number of unique for the individual row.) The first column will be the list of unique values; this will be the same for all of the arrays. The second column will be the counts for the corresponding values.
This will work even if the values stored in the array are not positive integers. It will even work if the values are not integers; however in the way it is written, it will consider two values to be different if they differ in even a single bit (so if the values were calculated and there might be round-off, there could be a problem.)
If you know that the values are all positive integers in the range 1 to something (such as 50) then the code can be simplified.
Best Answer