NumPy — From Basics to Advanced

A structured, well-commented Python reference covering NumPy from the ground up.
Each file is self-contained and executable — run it and read the output alongside the code.

---

Files

File	Level	Topics Covered
01_numpy_basics.py	Beginner	Array creation, dtypes, indexing, slicing, element-wise math, aggregations, shape manipulation, stacking, sorting, utilities
02_numpy_intermediate.py	Intermediate	Broadcasting, np.newaxis, fancy indexing, np.ix_, advanced sorting (lexsort, partition), set operations, linear algebra (linalg), statistics, structured arrays, masked arrays, polynomials, datetime, saving/loading
03_numpy_advanced.py	Advanced	Memory layout & strides, as_strided sliding windows, views vs copies, einsum, custom ufuncs, advanced broadcasting patterns, performance & memory tricks, memory-mapped files, FFT, advanced linear algebra (QR, Cholesky, SVD, lstsq), financial analytics, Numba teaser, common gotchas

---

How to Use

Each file is a runnable script with rich print output and inline explanations.

python 01_numpy_basics.py
python 02_numpy_intermediate.py
python 03_numpy_advanced.py

No setup beyond NumPy itself is required.

---

Requirements

numpy >= 1.20
python >= 3.8

Install NumPy if you don't have it:

pip install numpy

---

What You'll Learn

Basics

• Why NumPy is faster than Python lists (vectorisation benchmark included)
• Creating arrays with np.array, np.zeros, np.ones, np.arange, np.linspace
• Understanding shape, ndim, size, dtype, itemsize
• Indexing and slicing in 1D and 2D (including the view vs copy distinction)
• Element-wise math and universal functions (np.sin, np.exp, np.sqrt, ...)
• Boolean indexing and conditional selection
• Aggregations: sum, mean, std, min, max, argmin, argmax
• Shape manipulation: reshape, flatten, ravel, transpose
• Stacking (hstack, vstack) and splitting (hsplit, vsplit)

Intermediate

• Broadcasting rules and practical patterns (column/row normalisation, outer products)
• np.newaxis and expand_dims for dimension management
• Fancy indexing with integer arrays and np.where
• Deep-dive sorting: argsort, lexsort, partition
• Set operations: intersect1d, union1d, setdiff1d, isin
• Linear algebra: matrix multiply (@), det, inv, solve, eig, SVD, norms
• Statistics: percentiles, covariance, correlation, histograms
• Modern random API: np.random.default_rng
• Structured arrays (typed, named fields — like a lightweight table)
• Masked arrays with numpy.ma for missing-data handling
• Saving and loading with .npy, .npz, and .csv

Advanced

• C-order vs Fortran-order memory layout and cache performance
• Strides: how NumPy really works under the hood
• as_strided for zero-copy sliding windows (1D and 2D)
• Complete view vs copy truth table (shares_memory)
• einsum — matrix multiply, dot, outer, trace, batch ops, and more
• Ufunc internals: .reduce, .accumulate, .outer
• np.vectorize, np.frompyfunc, np.piecewise, apply_along_axis
• Advanced broadcasting: distance matrices, softmax, one-hot encoding
• Performance: out= parameter, keepdims, dtype choice, cache-friendly indexing
• Memory-mapped arrays (np.memmap) for out-of-RAM datasets
• FFT: signal decomposition, dominant frequency detection, 2D FFT
• Advanced decompositions: QR, Cholesky, pseudo-inverse, least-squares, eigh
• Financial analytics: log returns, Sharpe ratio, max drawdown, VaR
• Six common gotchas: integer overflow, float equality, view mutation, asarray vs array, chained indexing, keepdims

---

Structure

numpy/
├── 01_numpy_basics.py
├── 02_numpy_intermediate.py
├── 03_numpy_advanced.py
└── README.md

---

License

MIT