Optimise comparing multidimensional matrices

Hi. I'm trying to see where in a cell array of 21449x15, in which each cell contains a matrix of 1x7, each line of a matrix of 116255x7 occurs. Below is what I have as nested loops. It works as expected, however is obviously very slow. I'm struggling to apply any common optimisation techniques such as vectorisation or creating functions to this multidimensional problem. Would anyone have any advice please?

for i = 1:21449
    % For each formula from b1 to b15
    for j = 1:15
        % Skip empty cases
        if ~any(pep{i,j}) 
            continue
        else
            % For every formula in the MFP output
            for k = 1:length(formulas)
                % Check if the same formula exists in the database
                if  isequal(formulas, pep{i,j})
                    % If match found, replace the formula by a score, the
                    % higher the number in the sequence, the higher the
                    % score.
                    result(i,j) = 1*j;
                else
                    continue
                end
            end
        end
    end            
end

[Voss]

Can you upload a smaller version of your variables, say the first few hundred rows (or whatever makes sense) of pep and formulas, saved into a mat file (or two mat files)?

Also, is this line accurate as written?

if  isequal(formulas, pep{i,j})

Or is it actually this?

if  isequal(formulas(k,:), pep{i,j})

In any case, it would help to see (part of) the variables.

[Anna]

Thank you, that's correct, it should read formulas(k,:), I'm sorry. I have accepted the solution posted below.

Answer 1

[Bruno Luong]

I simply code by guessing what you have behind de scene. You better tell us exactly what are class/size of the variables in your pseudo code, for example the k loop doesn't oddly extract any elements from "formulas".

result = zeros(size(pep));
database = cat(1,pep{:});
tf = ismember(database, formulas, 'rows');
tf = reshape(tf, size(pep));
[~,j] = find(tf);
result(tf) = 1*j;

[Anna]

First of all I apologise for the quality of my question, I must have accidentally deleted the (k,:), it should read formulas(k,:).

Second of all, your solution is exactly what I was after! Thank you very much. Reshaping the array didn't even occur to me, that's very clever.