Find a row in a matrix in faster way

Hi,

I have to compare two matrices: in detail, I have to find if each row of a matrix1 is contained in a matrix2. Then I have to store the corresponding positions in a vector.

The following is a trivial method and seems to be faster than ismember version.

if true
    t1=tic;
    for m=1:size(matrix1,1)
      for l=1:size(matrix2,1)
        if(sum(xor(matrix1(m,:), matrix2(l,:)))==0)
            vector(m)=l;
        end   
      end
    end
   toc(t1)
end

ismember version.

if true
  t2=tic;
  for m=1:size(matrix1,1)
    index=ismember(matrix2,matrix1(m,:),'rows');
    l=find(index);
    vector(m)=l;
  end
  toc(t2)
end

Is there a way to speed up such computation? Thank

[Image Analyst]

Are they integer or floating point? http://matlab.wikia.com/wiki/FAQ#Why_is_0.3_-_0.2_-_0.1_.28or_similar.29_not_equal_to_zero.3F

[Matt Fig]

IA, the methods do not return the same results unless both matrices are binary....

Answer 1

[Matt Fig]

Here is a faster method, in the script I use to compare:

M1 = rand(1000,8)>.5;
M2 = rand(1000,8)>.5;

tic  % Method 1, use XOR
V1 = zeros(1,size(M1,1));
for m=1:size(M1,1)
    for n=1:size(M2,1)
        if(sum(xor(M1(m,:), M2(n,:)))==0)
            V1(m)=n;
        end
    end
end
toc

tic  % Method 2, use ISMEMBER
V2 = zeros(1,size(M1,1));
for m=1:size(M1,1)
    index=ismember(M2,M1(m,:),'rows');
    n = find(index,1,'last');
    if ~isempty(n)
        V2(m)=n;
    end
end
toc

tic  % Method 3, use BSXFUN
V3 = zeros(1,size(M1,1));
for m=1:size(M1,1)
    index=all(bsxfun(@eq,M1(m,:),M2),2);
    n = find(index,1,'last');
    if ~isempty(n)
        V3(m)=n;
    end
end
toc

isequal(V1,V2,V3)

[Matt Fig]

nedo nodo comments:

Hi,

thank you for your answer. I forgot to say that the matrices are binary I don't know if it makes some difference.

Thank you

[Matt Fig]

Did you see that I used binary matrices??

[nedo nodo]

Yes, thank you