Fastest way to find unique cells in a logical cell array

3 views (last 30 days)
Hello all,
I have a large array of cells (currently 1x4096 but likely to far bigger than that) with each cell of 6x6 logical values. Each cells contains 6 values that are 1 and rest are 0s. Only the placement of these 1s and 0s is different.
I want to find the fastest way to find unique cells in this array.
Currently I am using a for loop for finding unique cells like this:
for idx2 = 1:length(A)-1
for k = idx2 + 1 : length(A)
if(A{idx2} == A{k})
matching_indx(cnt) = uint16(k);
cnt=cnt+1;
end
end
end
But its really slow and does not scale well to larger cell arrays with larger values in each cell (8x8).
What would be the fastest way to achieve this?
Also what would be the most memory efficient way for these type of values? Would uisng sparse help in saving memory AND computations?
Regards

Accepted Answer

Guillaume
Guillaume on 7 Mar 2020
Edited: Guillaume on 7 Mar 2020
Why are you using a cell array to start with? A 6x6x4096 matrix would be more efficient in term of speed and memory. It also makes finding the histogram of your 6x6 array dead easy:
%converting cell array into N x N x numel(A) matrix
m = cat(3, A{:}); %better variable names than A and m required!
%finding unique NxN matrices and their count:
masrows = reshape(m, [], size(m, 3)).'; %a numel(A) x (NxN) matrix
[uniquem, ~, id] = unique(masrows, 'rows'); %find unique matrices and assign corresponding ID to each row
uniquem = reshape(uniquem.', size(m, 1), size(m, 2), []); %reshape into N x N x numuniquematrices
count = accumarray(id, 1); %one of the many ways to calculate histograms
If you're happy to store the matrices as a count x N x N array instead of a N x N x count array, you would avoid the two transpose above which would give you a speed gain.
Also, in matlab it's better to avoid using integer types unless you really need to.
  8 Comments
Haider Ali
Haider Ali on 8 Mar 2020
@Ameer Hamza, thank you! You have been really helpful.

Sign in to comment.

More Answers (0)

Categories

Find more on Creating and Concatenating Matrices in Help Center and File Exchange

Products


Release

R2019b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!