I have two arrays with shapes:
z1.shape
(74L, 1L)
z2.shape
(74L,)
why does on multiplication it produces 74x74 size array:
z3 = np.multiply(z1,z2)
z3.shape
(74L, 74L)
I was expecting an element by element multiplication to achieve z3 of shape (74L, 1L)
How do I achieve the z3 as element by element multiplication of z1 and z2
You can do z1*z2[:,None]. The extra None index stretches the 1D array into a column array.
Just do,
z3 = np.multiply(z1, z2.reshape(-1,1))
To answer your question, you need to reshape either array:
>>> np.multiply(z1.reshape(-1), z2).shape
(74,)
>>> np.multiply(z1, z2[:,np.newaxis]).shape
(74, 1)
Unless the shapes of z1, z2 are identical, the multiplication (or any function for that matter) will be cast as an outer product (z1 * z1.T), resulting in the shape (74,74) in your case. In the two examples above,
z1.reshape(-1) casts the first array to the shape of the second (74,); z2[:,np.newaxis] (or None instead of np.newaxis) casts the second to the shape of the first (74,1).
z1 * z2[None,:], a (74,1)*(1,74)