Replace rows of matrix with vector

So basically what I've done is create a function that allows me to replace a row within a matrix with a specified vector:

function [replacedM] = replace row(M,row,replace)
r=length(M);
for i=1:r
  if M(i,:)==row
    M(i,:)=replace;
  end
end
replacedM=M;
end

Now, I have 2 matrices about 13260X3 in size. One is called prechanged, the other is called changed (where row m of 'changed' is the modified row m of 'prechanged.') My main matrix, 'main,' is about about 76000X3 in size. Multiple rows of 'main' have triplicate rows.

For example, 'main' looks something like:

1 2 3 1
2 8 1 0
1 2 3 1
2 0 8 2
3 0 8 8
1 2 3 1
0 0 1 2

Rows 1, 3 , and 6 all have the same values.

'prechange' would look something like:

0 0 1 2
1 2 3 1
2 0 8 2
2 8 1 0
3 0 8 8

'changed' would look something like:

0 0 2 4
2 4 6 2
4 0 4 4
4 4 2 0
6 0 4 4

I'm implementing the replacerow function on 'main' like this:

for j=1:length(pre)
  main=replacerow(main,prechange(j,:),changed(j,:));
end

BUT it takes about 10 minutes (maybe because all main, prechange, and changed are extremely large matrices). Would anyone know if there is a way to shortening computational time? THANKS!

[Azzi Abdelmalek]

This is not clear, give a short example, and explain what you want

[Joseph Cheng]

yeah there are some inconsistencies. you say you have a Nx3 but your examples are Nx4.

[Azzi Abdelmalek]

How did you get 'changed' ?

[Jeremy]

My apologies. Just take out the last column. I made these matrices on the spot. So I have three matrices: main, prechanged, and changed. main is the original 76000X3 matrix with triplicates of rows, in no order. These rows represent vertices for strings of hexagonal shapes.

I used the unique function to remove the triplicates. With this new matrix of about 13260X3, I managed to group the 2210 hexagons, such that every group of 6 rows represented the vertices to a hexagon. This newly ordered matrix is prechanged. Then I found each hexagon's respective centroids in order to transform the hexagon by some factor k. (So each row of changed is just a modified row of prechanged - their row indices are the same if that makes sense.)

The problem is that I need to put these modified values back into the main matrix.

Answer 1

[Joseph Cheng]

well here is a thing that you could do to speed it up.

function [M] = replace(M,row,replace)
     [~,ind]=ismember(M,row,'rows');
     ind = find(ind==1);
     M(ind,:) = repmat(replace,length(ind),1);
end

and I tested it out using this

X = randi(10,76000,3);

prechange = randi(10,1000,3);
changed = randi(10,1000,3);
for i =1:length(prechange)
    rowsdupe = randi(76000,1,randi(10000,1,1));
    X(rowsdupe,:) = repmat(prechange(i,:),length(rowsdupe),1);
end

h=waitbar(0,'waiting')
tic
for j=1:length(prechange)
  X=replace(X,prechange(j,:),changed(j,:));
  waitbar(j/length(prechange),h,['performing step: ' num2str(j) '/' num2str(length(prechange))]);
end
close(h)
toc

elapsed time = 32.99 seconds.

[Jeremy]

Thanks! That seemed to help shorten the computational time by 2:30 minutes! I'll try to see what else might be causing the problem.