HN Showcase
Back to browse
GitHub Repository

Tiny ML & statistics library for JS — 16 algorithms in ~56KB gzipped. Rust/WASM, zero dependencies.

157 starsRust

8M algorithms in 56 KB – Rust/WASM library for JavaScript

by adam_perlinski·Feb 13, 2026·2 points·2 comments
Visit ProjectView on HN

AI Analysis

●●SolidSolve My ProblemNiche Gem

56KB ML library beats TensorFlow.js for forecasting and smoothing, but niche problem.

Strengths
  • •Aggressive size optimization: 56KB gzipped covers real use cases (trendlines, exponential fit, EMA)
  • •Rust/WASM foundation ensures sub-millisecond performance on typical datasets
  • •Live docs with runnable examples and clear real-world use cases (sales forecasts, sensor smoothing)
Weaknesses
  • •Limited to 16 algorithms—excludes clustering, classification, neural nets that competitors offer
  • •16 algorithms is narrow; scikit-learn, TensorFlow.js, and Danfo.js serve broader audiences
Target Audience

Frontend and Node.js developers needing lightweight ML for small datasets

Similar To

TensorFlow.js · Danfo.js · scikit-learn (Python equivalent)

Similar Projects