recommenders

Recommenders is an open-source project under the Linux Foundation of AI and Data that provides examples and best practices for building recommendation systems, delivered primarily as Jupyter notebooks. Its stated goal is to help researchers, developers and enthusiasts prototype, experiment with and bring classic and state-of-the-art recommendation algorithms toward production. The material is organized around five key tasks: preparing and loading data for each algorithm; building models with classical and deep learning approaches; evaluating algorithms with offline metrics; selecting and optimizing models via hyperparameter tuning; and operationalizing models in a production environment. Supporting utilities in the recommenders package handle common needs such as loading datasets in algorithm-specific formats, evaluating outputs and splitting training/test data. The repository documents a broad catalog of algorithms, including collaborative filtering methods such as ALS, BPR, BiVAE, Caser, LightGCN, NCF, RBM, SAR, SASRec, SVD, Wide and Deep, xDeepFM and various VAE variants, alongside content-based approaches such as DKN, LightGBM, LSTUR, NAML, NPA, NRMS, TF-IDF and Vowpal Wabbit. Each entry links to quick-start or deep-dive notebooks. A benchmark notebook compares several collaborative filtering algorithms on MovieLens 100k using ranking and rating metrics, with reported parameter settings and hardware context. Installation guidance recommends uv for environment management and VS Code for development, with optional extras for GPU, Spark, development and experimental features. The project is MIT licensed, welcomes contributions under published guidelines and a code of conduct, and cites related courses, books and academic papers.