pythonpython-3.xnumpycupy

How to multiply rows with corresponding columns in Python (cuPy/Numpy)?

I need to multiply each row with a column at the corresponding row-index. Consider the graphics below in which I have a 3 x 3 matrix. The required operation is to multiply the row[0] of matrix with col[0] of transposed_matrix, row[1] of matrix with col[1] of transposed_matrix, and so on.

enter image description here

Question: How can achieve it in cupPy/Numpy Python in a smart way (i.e. without using for-loops)?

Solution

  • Looks like you want:

    out = (A**2).sum(axis=1)

    Example:

    # input
array([[1, 2, 3],
       [4, 5, 6],
       [7, 8, 9]])

# output
array([ 14,  77, 194])

    Since you want the equivalent of sum(1*1 + 2*2 + 3*3) for each row.

    If A and B are different matrices, then use:

    out = (A*B.T).sum(axis=1)