libFM is a library for Factorization Machines (FM). FMs combine the advantages of predictors like linear/ polynomial regression or support vector machines (SVM) with factorization models. Like such models, FMs are a general predictor working with any real valued feature vector. In contrast to SVMs or polynomial regression, FMs model all interactions between variables using factorized parameters. Thus they are able to estimate higher order interactions even in problems with very large sparsity (like recommender systems) where SVMs fail. FMs can mimic several state-of-the-art factorization models including SVD++, PITF or FPMC just by specifying the input data.
This software implements the Pairwise Interaction Tensor Factorization (PITF) [WSDM 2010] model with BPR optimization for tag recommendation. It also contains a Factorization Machine implementation which can mimic PITF (see [ICDM 2010] for details).
MyMediaLite is a lightweight, multi-purpose library of recommender system algorithms. It addresses the two most common scenarios in collaborative filtering: rating prediction (e.g. on a scale of 1 to 5 stars), and item prediction from implicit feedback (e.g. from clicks or purchase actions). MyMediaLite was developed by Zeno Gantner, Steffen Rendle, and Christoph Freudenthaler.
Supplementary material & software
SIGIR 2011: Fast Context-aware Recommendations with Factorization Machines
- Multiverse Recommendation/ Tucker Decomposition
- Context-enrichment of Webscope dataset
- libFM with ALS optimization
ICDM 2010: Factorization Machines
- tag recommender with FM model
- libFM with SGD optimization