Fast context-aware recommendations with factorization machines

Steffen Rendle, Zeno Gantner, Christoph Freudenthaler, Lars Schmidt-Thieme

2011 · 591 citations · 20 references

Concepts

TL;DR

Context‑aware recommender systems use situational information to improve predictions, yet existing fast methods are specialized and lack generality. The state‑of‑the‑art approach, Multiverse Recommendation, employs Tucker tensor factorization to achieve high predictive accuracy. However, Multiverse Recommendation suffers from exponential complexity in the number of context variables and is restricted to categorical contexts.

Abstract

The situation in which a choice is made is an important information for recommender systems. Context-aware recommenders take this information into account to make predictions. So far, the best performing method for context-aware rating prediction in terms of predictive accuracy is Multiverse Recommendation based on the Tucker tensor factorization model. However this method has two drawbacks: (1) its model complexity is exponential in the number of context variables and polynomial in the size of the factorization and (2) it only works for categorical context variables. On the other hand there is a large variety of fast but specialized recommender methods which lack the generality of context-aware methods.

References

20