# ddbourgin/numpy-ml

Machine learning, in numpy

Repository: https://github.com/ddbourgin/numpy-ml
Canonical: https://ross.abutalabs.com/products/numpy-ml
Homepage: https://numpy-ml.readthedocs.io/
Language: Python
License: GPL-3.0
License Family: copyleft
Topics: machine-learning, neural-networks, topic-modeling, gaussian-mixture-models, hidden-markov-models, gradient-boosting, bayesian-inference, wavenet, vae, resnet, lstm, wgan-gp, attention, reinforcement-learning, good-turing-smoothing, knn, mfcc, gaussian-processes, word2vec
Last push: 2023-10-29T00:53:03+00:00

## Health v2 (maintenance only)
Score: 32/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 100
- inputs: {"age_days": 2706, "days_push": 1040, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 16330, forks 3742 (observed 2026-08-28T04:11:15.094706+00:00)

## What it is
numpy-ml is a collection of machine learning models and algorithms implemented exclusively in NumPy and the Python standard library, covering neural networks, HMMs, GMMs, topic models, tree-based models, and reinforcement learning. It serves as readable reference implementations for learning, prototyping, and experimentation rather than production use.

## Use cases
- learn how machine learning algorithms work by reading clean from-scratch implementations
- implement a neural network in pure numpy without deep learning frameworks
- study reference implementations of LSTM, attention, and ResNet blocks
- prototype ML experiments with minimal dependencies
- understand EM training for Gaussian mixture models and HMMs
- explore reinforcement learning agents on OpenAI gym environments
- implement topic modeling or word2vec from scratch

## When to choose
- you want legible, dependency-light implementations to read and modify
- you're teaching or learning ML fundamentals and want to see the math in code
- you need a starting point for rapid ML prototyping without heavy frameworks
- you want to benchmark your own implementation against a simple reference

## When to avoid
- you need production performance or GPU acceleration
- you're building a real application and should use PyTorch, TensorFlow, or scikit-learn
- you need well-tested, bug-free software - the project explicitly disclaims guarantees
- you need comprehensive documentation - docs are still under development

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, nlp, reinforcement-learning, data-science
- domain: machine-learning, deep-learning, education
- platform: python, cross-platform
- tags: numpy, reference-implementations, educational, from-scratch, neural-networks, topic-modeling, hidden-markov-models, gaussian-mixture-models, reinforcement-learning, no-framework-dependencies, natural-language-processing, algorithms

## Member repositories
- ddbourgin/numpy-ml (main) score 32

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:11:15.094706+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:05:08.934552+00:00, confidence not recorded.
  - readme: https://github.com/ddbourgin/numpy-ml (fetched 2026-08-28T04:11:15.094706+00:00, sha 8a9a2f21c535)
  - homepage: https://numpy-ml.readthedocs.io/ (fetched 2026-08-29T08:02:33.767772+00:00, sha 65edb51676b6)
  - registry_pypi: https://pypi.org/pypi/numpy-ml/json (fetched 2026-08-29T08:02:33.776740+00:00, sha efd2d55ee84c)
- Data as of 2026-08-30T08:39:29.467469+00:00.