The Ultimate NumPy Cheatsheet (Interactive)

This comprehensive cheat sheet provides interactive examples for learning NumPy, the fundamental Python library for numerical computing and array operations.

import numpy as np

The NumPy `ndarray` Object

The core of NumPy is the ndarray, an efficient, fixed-size, homogeneous array.

Key `ndarray` Attributes

type(data): Check the type of the array. Returns <class 'numpy.ndarray'>
data: View the array content (raw data buffer)
.ndim: The number of dimensions (axes). Returns integer value
.shape: A tuple indicating the size of the array in each dimension. Returns (rows, columns, ...)
.size: The total number of elements. Returns integer value
.dtype: The data type of the elements. Returns dtype object
.nbytes: The total bytes consumed by the array's elements. Returns integer value
.T: The transposed array (swaps dimensions)

Data Types and Type Casting

Specifying `dtype` on Creation

np.array(..., dtype=int): Create an array with integer type
np.array(..., dtype=float): Create an array with float type
np.array(..., dtype=complex): Create an array with complex type

Casting Data Types

np.array(data, dtype=int): Cast using np.array constructor
.astype(): Convert an array to a different data type

Type Promotion in Operations

Type Promotion: How types are promoted during operations (e.g., int + float = float)

Data Type Affecting Function Behavior

np.sqrt() with real numbers: Square root of negative numbers returns NaN
np.sqrt() with complex numbers: Square root with complex dtype handles negative numbers

Real and Imaginary Parts

.real: Get the real part of a complex array
.imag: Get the imaginary part of a complex array

Array Creation

From Existing Data & Sequences

np.array(1D list): Create a 1D array from a Python list
np.array(2D list): Create a 2D array from a list of lists
np.arange(): Create an array with regularly spaced values (step)
np.arange(float): Create an array with float step
np.linspace(): Create an array with a specific number of evenly spaced values
np.linspace(0, 10, 11): Example of linspace
np.logspace(): Create logarithmically spaced values

Initial Placeholders

np.zeros() / np.ones(): Create an array filled with zeros or ones
np.zeros(2x3): Example of np.zeros
np.ones(4): Example of np.ones
np.ones(dtype): np.ones with specified dtype
np.full() / np.empty(): Create a constant or uninitialized array
np.full() with value: Example of np.full
np.empty(): Example of np.empty
np.eye() / np.identity(): Create an identity matrix
np.identity(4): Example of np.identity
np.eye(k=1): Example of np.eye with offset
np.diag(): Create a diagonal array
np.random.*: Create arrays with random values

Meshgrid Arrays

np.meshgrid(): Create coordinate matrices from coordinate vectors

Indexing, Slicing, and Subsetting

1D Arrays Indexing & Slicing

np.arange(0, 11): Example array for 1D indexing
a[0]: Access the first element
a[-1]: Access the last element
a[4]: Access the fifth element
a[1:-1]: Slice from index 1 to -1
a[1:-1:2]: Slice with step 2
a[:5]: Slice up to index 5
a[-5:]: Slice from -5 to end
a[::-2]: Reverse slice with step 2

2D Arrays Indexing & Slicing

np.fromfunction(): Create a 2D array using a function
A[:, 1]: Select a column
A[1, :]: Select a row
A[:3, :3]: Select a block
A[::2, ::2]: Select every second element

Fancy and Boolean Indexing

Fancy Indexing (array): Using a NumPy array of indices
Fancy Indexing (list): Using a Python list of indices
Boolean Mask: Create a boolean mask
Boolean Filtering: Filter data using a boolean mask

Views vs. Copies

Understanding Views and Copies: Slices create views, while boolean/fancy indexing creates copies.

.copy(): Force the creation of a copy

Array Manipulation

Reshaping

.reshape(): Change the shape of an array without changing its data
np.reshape(data, (1, 4)): Example of np.reshape
.ravel() / .flatten(): Flatten an array to 1D (view vs. copy)
.flatten(): Example of .flatten()
np.newaxis (column): Add a new axis to create a column vector
np.newaxis (row): Add a new axis to create a row vector

Combining and Splitting

np.concatenate(): Join a sequence of arrays along an existing axis
np.vstack() / np.hstack(): Stack arrays vertically or horizontally
np.vstack(): Example of vertical stacking
np.hstack(): Example of horizontal stacking

Vectorized Expressions and Operations

Element-wise Operations (UFuncs)

Arithmetic: Standard operators (+, -, *, /, **) work element-wise
Element-wise Addition: Example of +
Element-wise Multiplication: Example of *
Element-wise Division: Example of /
Universal Functions: Functions like np.sqrt(), np.sin(), np.log()

Aggregate Functions

.sum() / .mean() / .std(): Compute sum, mean, standard deviation, etc.
np.mean(): Mean of random data
.mean(): Mean using method
.sum() (all): Sum of all elements
.sum(axis=0): Sum along columns
.sum(axis=1): Sum along rows
.min() / .max() / .argmin() / .argmax(): Find min/max values and their indices

Matrix and Vector Operations

@ or .dot(): Matrix multiplication
np.dot(A, x): Matrix-vector multiplication
A.dot(x): Matrix-vector multiplication (method)
np.dot(A, B): Matrix-matrix multiplication
np.dot(x, x): Inner product
np.outer(x, x): Outer product

Implementation Notes

This content was originally designed with interactive JavaScript modals for each clickable term. In the Markdown version, you can implement interactivity by:

Adding click handlers to code elements for showing explanations
Creating collapsible sections for detailed examples
Using JavaScript to generate dynamic examples
Implementing tooltips for inline explanations

Interactive Features (Future Implementation)

Click on any code element to see detailed explanation
Expandable sections for code examples
Live execution of NumPy operations
Step-by-step tutorials for complex operations

This NumPy cheatsheet covers the essential concepts and operations you'll need for scientific computing and data science applications.

Updated: January 15, 2025
Author: Danial Pahlavan
Category: Data Science & Scientific Computing