Easy way to do this in numpy?

Question

Suppose I have a BxNxL array, M. In other words, M is composed of B NxL matrices. In addition, I have a LxB column vector, Q. Is there any easy way (without for loops) to broadcast (sum) the ith column of Q to the ith matrix in M?

Answer 1

So your iterative code would be?

for i in range(...):    
    res[i,:,:] = M[i,:,:] + Q[:,i]    # NxL + L

with the whole array

res = M + Q.T[:,None,:]     # BxNxL  + (Bx1xL)

(I wrote this without a test example, so there might an error, but the basic idea should be right)