NumPy Cheat-Sheet

by

1. Basics

Importing NumPy

import numpy as np

Creating Arrays

arr = np.array([1, 2, 3])
matrix = np.array([[1, 2], [3, 4]])

Array Attributes

arr.shape # (3,)
arr.ndim  # 1
arr.size  # 3
arr.dtype # dtype('int64')

Array Initialization

zeros = np.zeros((2, 3))  # 2x3 matrix of zeros
ones = np.ones((3, 3))    # 3x3 matrix of ones
identity = np.eye(3)      # 3x3 Identity matrix
random = np.random.rand(2, 2)  # 2x2 matrix of random floats between 0 and 1

2. Array Indexing and Slicing

Indexing

arr[0]    # 1st element
matrix[1, 0]  # 2nd row, 1st column element

Slicing

arr[1:]   # [2, 3]
matrix[:, 1]  # 2nd column
matrix[1, :]  # 2nd row

Boolean Indexing/ Masking 

arr[arr > 1]  # [2, 3] (elements greater than 1)

3. Array Operations

Element-wise Operations 

arr + 5    # Add 5 to each element
arr * 2    # Multiply each element by 2

Matrix Multiplication 

matrix1 = [[1, 2], [2, 3]]
matrix2 = [[4, 5], [6, 7]]
np.dot(matrix1, matrix2)  # Matrix product
#So, A.B = [[1*4 + 2*6, 2*4 + 3*6], [1*5 + 2*7, 2*5 + 3*7]
#So the computed answer will be: [[16, 26], [19, 31]]

Broadcasting 

arr = np.array([1, 2, 3])
arr + np.array([1])    # Broadcast to [1, 1, 1]

Aggregate Functions 

np.sum(arr)       # Sum of all elements
np.mean(arr)      # Mean
np.min(arr)       # Minimum element
np.argmax(arr)    # Index of the max element

4. Reshaping and Transposing

Reshaping

reshaped = arr.reshape((3, 1))  # Convert 1D to 2D (3 rows, 1 column)

Transposing

matrix.T  # Transpose of the matrix

5. Advanced Indexing

  • Fancy Indexing
pythonCopy codearr = np.array([10, 20, 30, 40])
index = [0, 2]
arr[index]  # [10, 30]
  • Condition-based Indexing
pythonCopy codearr[arr > 15]  # [20, 30, 40]

6. Stacking and Splitting

  • Vertical and Horizontal Stacking
pythonCopy codea = np.array([1, 2, 3])
b = np.array([4, 5, 6])
np.vstack((a, b))  # Stack vertically
np.hstack((a, b))  # Stack horizontally
  • Splitting Arrays
pythonCopy codesplit_arr = np.split(arr, 2)  # Split array into 2 parts

7. Advanced Broadcasting

  • Broadcasting rules
pythonCopy codea = np.array([[1], [2], [3]])
b = np.array([4, 5, 6])
a + b  # Broadcast across dimensions

8. Linear Algebra

  • Determinant and Inverse
pythonCopy codenp.linalg.det(matrix)      # Determinant
np.linalg.inv(matrix)      # Inverse of the matrix
  • Eigenvalues and Eigenvectors
pythonCopy codeeigvals, eigvecs = np.linalg.eig(matrix)  # Eigenvalues and eigenvectors

9. Random Sampling

  • Random Numbers
pythonCopy coderand_arr = np.random.rand(3, 3)  # Random floats
rand_int = np.random.randint(1, 10, (3, 3))  # Random integers
  • Shuffling
pythonCopy codenp.random.shuffle(arr)   # Shuffle array in-place
  • Set Random Seed
pythonCopy codenp.random.seed(42)  # Set seed for reproducibility

10. Statistical Operations

  • Basic Statistics
pythonCopy codenp.mean(arr)  # Mean
np.median(arr)  # Median
np.std(arr)  # Standard deviation
  • Histogram
pythonCopy codehist, bin_edges = np.histogram(arr, bins=5)

11. Saving and Loading

  • Saving Arrays
pythonCopy codenp.save('array.npy', arr)  # Save as .npy file
np.savetxt('array.txt', arr, delimiter=',')  # Save as .txt
  • Loading Arrays
pythonCopy codearr = np.load('array.npy')
arr = np.loadtxt('array.txt', delimiter=',')

12. Broadcasting Rules

NumPy allows broadcasting of arrays for element-wise operations, provided:

  1. Dimensions are compatible (either equal or one is 1).
  2. The smaller array is stretched without copying data.

13. Advanced ufunc (Universal Functions)

  • Custom ufunc
pythonCopy codedef custom_function(x):
    return x * 2

ufunc = np.frompyfunc(custom_function, 1, 1)
ufunc(np.array([1, 2, 3]))  # Apply custom ufunc

This cheatsheet covers essential NumPy functionalities, from basic array manipulations to advanced topics like linear algebra and broadcasting.

Leave a Comment