how to multiply each row of a matrix with numpy-CodePudding

I am new on machine learning.Using python, numpy. I need to get a dot product on a matrix with size (3, 2) and each row of a matrix with size (100, 2), which is

a = [[1, 2], [3, 4], [5, 6]]
b = [[5, 5], [6, 6], [7, 7], ...] # it has 100 row

and what i want is:

np.dot(a, b[0])
np.dot(a, b[1])

i currently have:

z = np.dot(a, b)

but the dimension doesn't match

but i cannot use loop and the code need to be vectorized.

can anyone give me some hints, thank you so much!!

CodePudding user response：

It seems you want to compute the dot product of the matrix of shape 3,2 with the transpose of the matrix of shape 100,2

You can get the transpose of a matrix m with m.T

So what you want is:

np.dot(a, b.T)

That will give you a matrix of shape 3,100 where each column is np.dot(a,b[i]) for i = 0,...,99

CodePudding user response：

For a dot product to work, it requires the first dimension of b to match the second dimension of a. Thus you need to transpose b:

np.dot(a, b.T)

Output:

array([[15, 18, 21, 24],
       [35, 42, 49, 56],
       [55, 66, 77, 88]])

Or a, depending on the expected output:

>>> np.dot(b, a.T)
array([[15, 35, 55],
       [18, 42, 66],
       [21, 49, 77],
       [24, 56, 88]])