This project aims at understanding and replicating recent work that uses Bayesian methods in graph-based recommendation systems. We implement the paper 'Efficient Bayesian Methods in Graph-based Recommendation' by Lopes et al. We further propose three extensions to this paper that incorporate user-user and item-item similarities, as well as user reliability into the probabilistic ranking functions described in the original paper. The results are intuitive, and changes in performance can also be easily explained.
- This project was done as a part of the course CSE 291 "Recommender Systems" course (Fall 2017 quarter) at UC San Diego.