12 Ways to Boost Data Science Matrix Ops: Numpy to Cython

Python Matrix Ops: Faster, Further, Better

Photo by Bryanquille on Pixabay

Lately, I’ve been using Python for scientific computing, but I’m bothered by its slow speed.

Matrix calculations are also sluggish. Let me share some tips to speed things up.

You can accelerate in two ways: through high-level API and low-level API.

Low-Level API

Photo by Olivier Collet on Unsplash

The simplest and crudest way is to directly traverse with a triple loop:

def matrix_multiplication_loop(A, B):
    m = A.shape[0]
    n = A.shape[1]
    l = B.shape[1]
    C = np.zeros([m, l])
    for i in range(m):
        for j in range(l):
            for k in range(n):
                C[i][j] += A[i][k] * B[k][j]  # Accumulate the result of each multiplication
    return C
A = np.random.random([300, 12])
B = np.random.random([12, 256])
%timeit C = matrix_multiplication_loop(A, B)
1 loop, best of 3: 2.22 s per loop