# Distributed Bayesian Probabilistic Matrix Factorization

## TL;DR

This paper introduces a distributed, high-performance parallel implementation of Bayesian Probabilistic Matrix Factorization, significantly improving scalability and efficiency for large-scale recommender system data.

## Contribution

It presents a novel distributed and parallel implementation of BPMF that outperforms existing methods through load balancing and asynchronous communication.

## Key findings

- Achieved faster computation times than state-of-the-art implementations.
- Demonstrated scalability on shared memory and distributed architectures.
- Improved load balancing with work stealing enhances performance.

## Abstract

Matrix factorization is a common machine learning technique for recommender systems. Despite its high prediction accuracy, the Bayesian Probabilistic Matrix Factorization algorithm (BPMF) has not been widely used on large scale data because of its high computational cost. In this paper we propose a distributed high-performance parallel implementation of BPMF on shared memory and distributed architectures. We show by using efficient load balancing using work stealing on a single node, and by using asynchronous communication in the distributed version we beat state of the art implementations.