NumPy ML (numpy-ml) is a collection of machine learning algorithms implemented exclusively in NumPy: deliberately inefficient but highly legible code that shows exactly how each algorithm works. Where frameworks like PyTorch hide the math behind optimized kernels, numpy-ml spells it out, from neural networks (including RNNs, LSTMs, VAEs and WGAN-GP) to classical methods such as gradient boosting, Gaussian mixture models, hidden Markov models and Gaussian processes, plus reinforcement learning agents and NLP models like word2vec. The project has more than 16,000 stars on GitHub.
The library is built for learning and prototyping. You can install it from PyPI, clone it as a starting point for experiments, and read each model end to end because everything is plain NumPy with documented equations. Coverage is broad: supervised and unsupervised learning, topic models, attention mechanisms, tree ensembles, kernel methods and RL agents that train on OpenAI Gym environments. Full documentation lives on Read the Docs, with per-model details. The codebase was last updated in 2023, so treat it as a stable educational reference rather than a fast-moving framework, but for understanding the internals of ML algorithms it remains one of the clearest resources available.
NumPy ML is completely free and open source under the GPL-3.0 license. There are no paid tiers.
NumPy ML is for students, self-taught ML engineers and researchers who want to understand algorithms from the inside, and for anyone prototyping an idea who prefers transparent code they can modify line by line.