A Rust ML library that knew when to quit

What it does

rusty-machine is a general-purpose machine learning library written entirely in Rust, covering the usual suspects: linear and logistic regression, k-means, neural networks, SVMs, Gaussian processes, DBSCAN, PCA, and more. It bundles its own linear algebra via rulinalg and exposes models through train/predict traits for supervised and unsupervised learning.

The interesting bit

The author built this to learn Rust and relearn ML algorithms — then was honest enough to flag it as “probably not the best choice for any serious projects” and eventually abandon it entirely. That candor is rarer than a bug-free gradient descent. The zero-dependency stance (no BLAS, no LAPACK) keeps builds simple at the cost of leaving performance on the table.

Key highlights

• Covers a wide algorithmic spread for a solo/small project: GLMs, Gaussian mixture models, naive Bayes, k-NN, plus a basic stats module behind a feature flag
• Traits (SupModel, UnSupModel) enforce a consistent API across models
• Re-exports its linear algebra crate to avoid forcing users to juggle dependencies
• Version 0.5.4, frozen in amber; Travis CI badge still blinking in the afterlife

Caveats

• No longer maintained — the README’s first words, not a footnote
• No BLAS/LAPACK integration means matrix operations are pure Rust, likely slower than competitors
• Self-described as early-stage and incomplete; the author actively solicited contributors before stepping away

Verdict

Worth a look if you’re studying how to implement ML algorithms in Rust, or want a lightweight, dependency-free sandbox. Skip it for production work — the ecosystem has moved on, and so did the maintainer.