ismember matlab function too slow

Hi, I would love some help optimizing this piece of code.

In my model i have to go every time instant through all the pixels of an image ( this cannot be changed to solve the equations that I want).

At a given step I want to now if a given pixel corresponds to the list of upetake indexes (uptake_indices) or not. If so I would have a concentration = uptake, else it would be 0.

Currently I am using ismember, but since my arrays are big and the loops are really long, this step delays a lot my code.

Here is a pice of my code.

Thanks for the help!

% list of indexes of my 400x400 image
aux=1;
pareja_indice=zeros(400*400,1);
for i=1:400
    for j=1:400
        pareja_indice(aux) = sub2ind([400, 400], i, j);
        aux=aux+1;
    end
end

% Load indexes corresponding to uptake points in the image
load('GABA_uptake_points3.mat');
uptake_indices = sub2ind([400, 400], SourceY, SourceX);

aux=0;
uptake=30;

% initialize concentration matrix
C=zeros(400,400);

for k=1:2000000
 % other operations
    for i=1:400
        for j=1:400
            aux=aux+1;
            % we should only have uptake at the uptake locations
            C_up = uptake * ismember(pareja_indice(aux), uptake_indices);
            C(i,j)=C_up;
            % ... other operations 
        end
    end
% other operation
end

Answer 1

[Hitesh]

Hi Paula Giménez Mínguez,

From the code snippet, what I understand that you are determining if a pixel index matches an uptake index, which is loaded from the "GABA_uptake_points3.mat" file using the SourceY and SourceX variables. By transforming uptake_indices into a logical array, you can efficiently identify whether each pixel is an uptake point. This method enables direct indexing, offering a significant improvement in elapsed time over repeatedly using ismember.

Refer to the below code :

tic;
% List of indexes of the 400x400 image
pareja_indice = reshape(1:(400*400), [400, 400]);

% Load indexes corresponding to uptake points in the image
load('GABA_uptake_points3.mat');
uptake_indices = sub2ind([400, 400], SourceY, SourceX);

% Create a logical array to identify uptake points
uptake_mask = false(400, 400);
uptake_mask(uptake_indices) = true;

% Initialize concentration matrix
uptake = 30;
C = zeros(400, 400);

for k = 1:2000000
    % Only perform operations on uptake points
    C(uptake_mask) = uptake;

    % ... other operations
end
toc;

For more information on logical array, refer to this MATLAB documentation: https://www.mathworks.com/help/matlab/ref/logical.html

Hope this helps!!

[Paula Giménez Mínguez]

Thank you so much!

Indeed, finally I did something very similar and it reduced importantly the elapse time. What it took days of runing now is just a couple of hours.