Fast Non-Bayesian Poisson Factorization for Implicit-Feedback Recommendations
This provides faster and more accurate recommendations for users in implicit-feedback systems, though it is incremental as it builds on existing Poisson factorization methods.
The paper tackles the problem of slow Bayesian Poisson factorization for implicit-feedback recommendations by proposing a non-Bayesian optimization-based approach, resulting in better top-N recommendations with significantly shorter fitting times and sparse solutions.
This work explores non-negative low-rank matrix factorization based on regularized Poisson models (PF or "Poisson factorization" for short) for recommender systems with implicit-feedback data. The properties of Poisson likelihood allow a shortcut for very fast computations over zero-valued inputs, and oftentimes results in very sparse factors for both users and items. Compared to HPF (a popular Bayesian formulation of the problem with hierarchical priors), the frequentist optimization-based approach presented here tends to produce better top-N recommendations with significantly shorter fitting times, on top of having sparse solutions.