Turbo-CF: Matrix Decomposition-Free Graph Filtering for Fast Recommendation

TL;DR

Turbo-CF is a fast, training-free graph filtering method for recommendation systems that avoids matrix decomposition, enabling rapid and accurate recommendations suitable for real-time applications.

Contribution

It introduces a matrix decomposition-free, polynomial graph filter approach that significantly speeds up collaborative filtering without sacrificing accuracy.

Findings

Runtime under 1 second on benchmark datasets

Achieves comparable accuracy to state-of-the-art methods

Utilizes GPU effectively for fast processing

Abstract

A series of graph filtering (GF)-based collaborative filtering (CF) showcases state-of-the-art performance on the recommendation accuracy by using a low-pass filter (LPF) without a training process. However, conventional GF-based CF approaches mostly perform matrix decomposition on the item-item similarity graph to realize the ideal LPF, which results in a non-trivial computational cost and thus makes them less practical in scenarios where rapid recommendations are essential. In this paper, we propose Turbo-CF, a GF-based CF method that is both training-free and matrix decomposition-free. Turbo-CF employs a polynomial graph filter to circumvent the issue of expensive matrix decompositions, enabling us to make full use of modern computer hardware components (i.e., GPU). Specifically, Turbo-CF first constructs an item-item similarity graph whose edge weights are effectively regulated. Then, our own polynomial LPFs are designed to retain only low-frequency signals without explicit matrix decompositions. We demonstrate that Turbo-CF is extremely fast yet accurate, achieving a runtime of less than 1 second on real-world benchmark datasets while achieving recommendation accuracies comparable to best competitors.