I have a large two dimentional numpy array A of dimension around 250000 x 30 and two one dimensional numpy arrays x and y. I want to extract the sub-array of A with rows in x and columns in y. Which method is more efficient?
A[x[:,np.newaxis], y]A[np.ix_(x,y)]Benchmark it!
import numpy as np
# some data
A = np.random.random((250000, 30))
# some random indices
x = np.random.randint(0, 250000, 150000)
y = np.random.randint(0, 30, 10)
def method1(A, x, y):
return A[x[:, np.newaxis], y]
def method2(A, x, y):
return A[np.ix_(x,y)]
def method3(A, x, y):
return A[x][:,y]
def method4(A, x, y):
return A[:,y][x]
These three methods give the following benchmarks:
method1: 87.7 ms
method2: 89.2 ms
method3: 115 ms
method4: 141 ms
So, the answer is that there is not real difference between the two methods in the question.
