how to find euclidean distance between training and test data?

I have two matrices A of size 2x5 and B of size 2x2 such that each column is a feature vector. I want to calculate the euclidean distance between A and B. i.e I want to calculate the euclidean distance between first column of B with every column of A and similarly need to calculate the second column of B with every column of A .

   A=rand(5,2);
   B=rand(2,2);
      for i=1:2
        for j=1:5
         d=sqrt(B(i,:)-A(j,:));
    end

end

Answer 1

[Walter Roberson]

Statistics toolbox:

 d = pdist2(A.', B.');

pdist2 normally operates row by row, which is why the .' are there, to make your column-oriented data into row-oriented data.

[kitty varghese]

Im getting an error.

Error using pdist2 (line 138)

X and Y must have the same number of columns.

[Walter Roberson]

Your problem statement says,

I have two matrices A of size 2x5 and B of size 2x2

and that

each column is a feature vector.

But your sample code has

   A=rand(5,2);

which is a 5 x 2 array, not a 2 x 5 array. The columns of your actual A are length 5, and the columns of your actual B are length 2, and it is not meaningful to find the euclidean distance between objects with different lengths.

Your sample code has

sqrt(B(i,:)-A(j,:))

which is comparing all of the columns of particular rows, which would be suitable for the case where the rows are feature vectors, not column vectors.

If your rows are feature vectors then pdist2(A, B)