Collaborative Diffusion Model for Recommender System

TL;DR

This paper introduces CDiff4Rec, a collaborative diffusion model that enhances recommender systems by integrating item content and collaborative signals, overcoming limitations of existing diffusion-based recommenders.

Contribution

The paper proposes a novel diffusion-based recommender system that generates pseudo-users from item features and leverages collaborative signals to improve personalization.

Findings

Outperforms existing methods on three public datasets.

Effectively mitigates loss of personalized information.

Integrates item content with collaborative signals for better recommendations.

Abstract

Diffusion-based recommender systems (DR) have gained increasing attention for their advanced generative and denoising capabilities. However, existing DR face two central limitations: (i) a trade-off between enhancing generative capacity via noise injection and retaining the loss of personalized information. (ii) the underutilization of rich item-side information. To address these challenges, we present a Collaborative Diffusion model for Recommender System (CDiff4Rec). Specifically, CDiff4Rec generates pseudo-users from item features and leverages collaborative signals from both real and pseudo personalized neighbors identified through behavioral similarity, thereby effectively reconstructing nuanced user preferences. Experimental results on three public datasets show that CDiff4Rec outperforms competitors by effectively mitigating the loss of personalized information through the integration of item content and collaborative signals.