NumPy Reference

Array Operations

import numpy as np

matrix = np.zeros([rows, columns])
matrix[2, 3]  # access row 2, column 3

Common types and functions:

• np.uint8, np.uint16, np.float32 — common image data types

• np.around(arr) — element-wise rounding

Log-Sum-Exp

For numerically stable sum of log probabilities:

from scipy.special import logsumexp

logsumexp([-2, -3])  # = np.log(np.exp(-2) + np.exp(-3))
# logsumexp(a, b) = log(e^a + e^b)

Resources