Comparing values across unequal matrices in Matlab - matlab

Question

I would like to compare rows across two unequal matrices in Matlab and extract these rows to be stored in a different matrix (say D). For example,
tmp = [2245; 2345; 2290; 4576]
and
id=[1 2245 564 8890 123;
2 2445 5673 7846 342;
3 2290 3428 3321 908].
Id is a much larger matrix. I want to locate each value of tmp which is in ‘id’. Although using the intersect command in the line below I have been able to locate the rows of id which contain the values from tmp, I would like to do this for each value of tmp one by one as each value of tmp is repeated multiple times in id. I tried using foreach. However, I get an error message stating that foreach cannot be used for char type array. Could anyone please suggest an alternative how to go about this?
for j=1:length(tmp);
[D,itmp,id2] = intersect(tmp(j,1),id(:,2), 'rows');
Despite using the loop, the code doesn’t seem to take one value of j at a time. This was the reason behind trying ‘foreach j’. Also after finding the rows common to the two matrices and storing them in D, I would like to append matrix id to include the value of j next to the relevant row within id. For example, if the first value within tmp was repeated in id in rows 1,3,5,10; I would like a column in id which would take the value 1 next to rows 1,3,5,10. Any help on this would be much appreciated! Thanks.

Answer 1

Not sure exactly, what you trying to do, but to search a value in a matrix you can use find:
for i = 1:numel(tmp)
[row, col] = find(id == tmp(i));
end

Answer 2

You can easily achieve this using a combination of bsxfun and permute. What you would do is transform the tmp vector so that it is a single 3D vector, then use the eq (equals) function and see which values in your matrix are equal to each value of tmp. Therefore, do something like:
%// Your data
id=[1 2245 564 8890 123;
2 2445 5673 7846 342;
3 2290 3428 3321 908]
tmp = [2245; 2345; 2290; 4576];
tmp2 = permute(tmp, [3 2 1]); %// Make a 3D vector
tmp3 = bsxfun(#eq, id, tmp2); %// Find which locations of id are equal to each value of tmp
This is what I get for my final output, stored in tmp3:
tmp3(:,:,1) =
0 1 0 0 0
0 0 0 0 0
0 0 0 0 0
tmp3(:,:,2) =
0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
tmp3(:,:,3) =
0 0 0 0 0
0 0 0 0 0
0 1 0 0 0
tmp3(:,:,4) =
0 0 0 0 0
0 0 0 0 0
0 0 0 0 0
As you can see here, each 3D slice tells you which elements in id match the corresponding value in tmp. Therefore, the first slice tells you whether any element in id is equal to tmp(1), which is 2245. In this case, this would be the first row and second column. For the second slice, no values matched tmp(2) = 2345. For the third slice, one value in id matched tmp(3) = 2290, which is row 3, column 3.
Now, what you're really after is determining the rows and columns of where each value of tmp matched each location in id. You want this to be delineated per id number. That can easily be done using ind2sub and find on this matrix:
[rows, cols, ID] = ind2sub(size(tmp3), find(tmp3))
rows =
1
3
cols =
2
2
ID =
1
3
Therefore, ID tells you which id we have matched to, and rows, cols tells you which rows and columns in id we were able to match with. Therefore, for id = 1, we found a match with tmp(1)= 2245, and this is located at row=1,col=2. Similarly, for id = 3, we found a match with tmp(3)=2290 and this is at row=3,col=2.
To make this into all one big 2D matrix that contains all the information you want, you can simply concatenate all of these columns into one matrix. Therefore:
final = [ID rows cols]
final =
1 1 2
3 3 2
You can read this like so:
id=1 with tmp(1) = 2245, we found this value in row=1,col=2.
id=3, with tmp(3) = 2290, we found this value in row=3,col=2
Hope this helps!