Compare matrix column-wise with vector

Dear all,

I have an mxn matrix and a 1xn vector I want to compare each ith column of the matrix with the ith value of the vector. The result should be an mxn logical matrix to be used

Example:

minVec = [ 0  2  0];

maxVec = [10 10  9];

mat    = [-1  2  5;
           9 15  3];

I would like to calculate the following, but without a for-loop...

mask(:,i) = mat(:,i)<minVec(i) | mat(:,i)>maxVec(i);

mask   = [ 1  0  0;
           0  1  0]

I guess there must be some solution using arrayfun, perhaps?

Any help is greatly appreciated!

Thank you very much, Philip

Answer 1

[Azzi Abdelmalek]

mask=bsxfun(@lt,mat,minVec)| bsxfun(@gt,mat,maxVec)

[Philip Ohnewein]

Oh, wow, what an elegant, simple and fast solution: works like a charm, thank you very much! I didn't even know that a function called bsxfun exists... Well, I'm a newbie, after all... :-)

I have another somehow similar problem, maybe this can be solved in a similar way? Here it is:

Given mx1 vector 'vec' and 1xn vector 'threshholds', I want to construct an 1xn vector 'minIdx': minIdx(i) should contain the index of the element of vec that is closest to threshholds(i), but greater than thresh(i). Example:

vec = [3 4 5 6 7 8]';

threshholds = [4.5 3];

minIdx = [3 2]

I guess this is possible again using bsxfun or arrayfun or something like that. I am currently doing it with some repmats, but I don't really like that...

Thank you very much!

Answer 2

[Image Analyst]

Try this:

[rows, columns] = size(mat)
mask = mat < repmat(minVec, [rows, 1]) | mat > repmat(maxVec, [rows, 1])

[Philip Ohnewein]

Thank you, this works. I was actually using a similar code, but needed a better solution, because the use of repmat is both slow and very memory hungry, since mat has a huge number of rows, about 10^6.

That's why I was looking for some more efficient code.