NumPy

NumPy provides multi-dimensional arrays and a library of mathematical operations that run in optimised C and Fortran code. It is the common data exchange format of the scientific Python stack — Pandas, SciPy, scikit-learn, PyTorch, and TensorFlow all interoperate with NumPy arrays. Vectorised operations over arrays replace explicit Python loops, giving 10–100× speedups for numerical work.

Quick start

pip install numpy

import numpy as np

# Create arrays
a = np.array([1, 2, 3, 4, 5])
b = np.arange(1, 6)              # [1, 2, 3, 4, 5]
c = np.zeros((3, 4))             # 3x4 matrix of zeros
d = np.random.rand(3, 3)         # 3x3 matrix of random floats

# Vectorised operations (no loop needed)
print(a * 2)                     # [2, 4, 6, 8, 10]
print(a + b)                     # [2, 4, 6, 8, 10]
print(a.mean(), a.std())

# 2D array operations
matrix = np.array([[1, 2], [3, 4]])
print(matrix.T)                  # transpose
print(np.linalg.det(matrix))     # determinant: -2.0
print(np.linalg.inv(matrix))     # inverse

# Broadcasting
x = np.array([[1, 2, 3],
               [4, 5, 6]])
print(x + np.array([10, 20, 30]))  # add row-wise

# Boolean indexing
scores = np.array([72, 45, 88, 91, 60])
print(scores[scores > 70])          # [72, 88, 91]

When to use

NumPy is essential for any Python numerical or scientific computing work. It’s a dependency of virtually every data science library and you’ll use it even when working at higher levels (Pandas, PyTorch). Use NumPy directly for linear algebra, signal processing, and custom numerical algorithms. For tabular data and time series, Pandas adds labelled indexing on top. For GPU-accelerated array computing, CuPy provides a NumPy-compatible API.