Accelerating Recommender Systems using GPUs

TL;DR

This paper presents GPU implementations of matrix recommender algorithms CCD++ and ALS, demonstrating significant speedups over multi-core versions while maintaining comparable predictive accuracy.

Contribution

The paper introduces GPU-based implementations of CCD++ and ALS algorithms for recommender systems, achieving up to 14.8x speedup over multi-core counterparts.

Findings

GPU implementations outperform multi-core versions in processing time

Maximum speedup achieved is 14.8 times

Predictive accuracy remains comparable to existing methods

Abstract

We describe GPU implementations of the matrix recommender algorithms CCD++ and ALS. We compare the processing time and predictive ability of the GPU implementations with existing multi-core versions of the same algorithms. Results on the GPU are better than the results of the multi-core versions (maximum speedup of 14.8).