Subtracting one column in matrix 1 from each column in matrix 2

Hi, I have to matrix (M_1, size 90x31 and M_2, size 90x130), i would like to subtract one column in matrix M_1 from each column in matrix M_2 and repeat this for each columns in matrix M_1 (31 times), and the result is a matrix ( there are 31 rows and 130 columns. Any ideas?

Thank you in advance.

[Voss]

Subtracting two column vectors results in another column vector of the same size - in this case 90-by-1.

So the idea here is to do this type of operation 130*31 times, resulting in 130*31 column vectors of size 90-by-1. These will not all 'fit' in a 31-by-130 matrix; you need another dimension of size 90. That is, you will end up with a 90-by-31-by-130 matrix or a 31-by-130-by-90 matrix, etc.

On the other hand, if somehow you define the process of subtracting two column vectors to result in a scalar value, then you can end up with a 31-by-130 matrix.

So, you can definitely do what you want, but it'll give you a 3d matrix. e.g., 31-by-130-by-90. Is that ok for your purposes?

[Pawel Szczepanowski]

You're right, i made mistake when i wrote my problem. I mean that i have to calculate Euclidean distance between first column in martix M_1 from each columns in martix M_2 (the result is a vector) and repeat this caluclation to all the rest of the columns in matrix M_1, so it'll give me 2d matrix with 31 rows. I have no idea how to calculate Euclidean distance between first column in martix M_1 from each columns in martix M_2. Do u have any idea?

Thank u in advance.

[Image Analyst]

What do the rows and columns represent? Do you have a 90 or 130 dimensional space? Or is each column just the location, like the x value along a number line? Have you seen pdist2() in the Stats toolbox?

[Pawel Szczepanowski]

each column in M_1 is a column vector of picture (picture is reshaped) is the same with martix M_2. I'd like to calculate euclideus distance between first column vector form M_1 and all column vectors in M_2 and repeat it with all column in M_1.

Answer 1

[Torsten]

Dist = vecnorm(bsxfun(@minus,M_2,M_1(:,1)));
for i=2:31
    Dist = vertcat(Dist,vecnorm(bsxfun(@minus,M_2,M_1(:,i))));
end

[Pawel Szczepanowski]

Thank you very much!