MLAlgorithms (rushter)
A collection of minimal from-scratch Python implementations of classical machine-learning algorithms, written on NumPy/SciPy/Autograd for readability rather than speed. MIT-licensed, ~11.1k stars / 1.8k forks, by the GitHub user rushter. Its stated audience is people who want to learn the internals of ML algorithms or implement them themselves.
What it covers
Linear and logistic regression · SVM with linear/polynomial/RBF kernels · decision trees · random forests · gradient boosting · k-means · Gaussian mixture models · kNN · naive Bayes · PCA · t-SNE · factorization machines · restricted Boltzmann machines · neural networks (MLP/CNN/RNN/LSTM) · deep Q-learning.
That span is the point: it crosses the classical and neural eras in one consistent style, so the same reader can see a decision tree and an LSTM written to the same standard.
Read it as a reference, not a library
151 commits and stable — a fixed teaching corpus rather than a maintained dependency. Nothing here is meant for production; the implementations trade speed for legibility, which is exactly the trade that makes them readable and exactly why they don’t belong in a serving path.
Tier
T2. A primary artifact — the code is the claim, and it is directly inspectable. Not T1: it is one developer’s reference implementation with no accompanying derivation, benchmark, or peer review, so it establishes how an algorithm can be written, not that it is written correctly or optimally.
Why it’s here
It anchors the layer this wiki’s other founding sources skip. finetuning-explained-kdnuggets and unsloth-amd-support start from a pretrained model and adapt it; this one starts from the fitting procedures underneath. See classical-ml-algorithms and machine-learning.
The hub router twice declined to file it elsewhere: ../optimization-algorithms-wiki owns
metaheuristic global optimizers and would have conflated two meanings of “optimization,” and the
parked learning-roadmaps cluster is curricula — meta-documents pointing at other people’s
material — where this is the material.