xlearn

Why we included this project

xLearn is for anyone wrestling with high-dimensional sparse data, the kind you get from recommendation systems with millions of users and item categories, or click-through prediction on raw categorical features. It wraps linear models, factorization machines, and field-aware factorization machines behind a simple Python API and a CLI, so you can prototype in a notebook and later run the same model at scale without rewriting code. The C++ core is tuned for cache behavior and lock-free updates, which shows up as real speed gains on large sparse inputs, and out-of-core training streams from disk, letting you fit datasets bigger than RAM. Built-in cross-validation and early stopping round it out as a practical tool for structured-data modeling rather than a research toy.