Recommender Systems
How products suggest what you will like: collaborative filtering, hybrid models, ranking and evaluation.
Advanced
Once the basics are comfortable.
Implicit: Fast Collaborative Filtering for Implicit Datasets
A fast Python library for recommending from clicks, plays and purchases instead of star ratings, with alternating least squares and Bayesian personalised ranking.
Sign up to see the source
LightFM: Hybrid Recommender
A library that mixes collaborative filtering with item and user features, which helps with new users and new items. The repository has a quickstart and a worked example.
Sign up to see the source
Recommenders: Best Practices on Recommendation Systems
A large collection of notebooks that take a recommender from data preparation through many algorithms and evaluation metrics to deployment.
Sign up to see the source
System Design for Recommendations and Search
How real products structure recommendation and search: offline and online parts, candidate retrieval, ranking and feedback loops, compared across companies.
Sign up to see the source
TensorFlow Recommenders
Tutorials for building a full recommender in two stages, retrieval then ranking, using deep learning, side features and sequence models, on real movie-rating data.
Sign up to see the source