Denoising Diffusion Recommender Model

TL;DR

DDRM is proposed, which leverages multi-step denoising process of diffusion models to robustify user and item embeddings from any recommender models and proposes a dedicated denoising module that encodes collaborative information as denoising guidance.

Abstract

Recommender systems often grapple with noisy implicit feedback. Most studies alleviate the noise issues from data cleaning perspective such as data resampling and reweighting, but they are constrained by heuristic assumptions. Another denoising avenue is from model perspective, which proactively injects noises into user-item interactions and enhances the intrinsic denoising ability of models. However, this kind of denoising process poses significant challenges to the recommender model's representation capacity to capture noise patterns. To address this issue, we propose Denoising Diffusion Recommender Model (DDRM), which leverages multi-step denoising process of diffusion models to robustify user and item embeddings from any recommender models. DDRM injects controlled Gaussian noises in the forward process and iteratively removes noises in the reverse denoising process, thereby improving embedding robustness against noisy feedback. To achieve this target, the key lies in offering appropriate guidance to steer the reverse denoising process and providing a proper starting point to start the forward-reverse process during inference. In particular, we propose a dedicated denoising module that encodes collaborative information as denoising guidance. Besides, in the inference stage, DDRM utilizes the average embeddings of users' historically liked items as the starting point rather than using pure noise since pure noise lacks personalization, which increases the difficulty of the denoising process. Extensive experiments on three datasets with three representative backend recommender models demonstrate the effectiveness of DDRM.

Figures (7)

• Figure 1: Illustration of resampling, reweighting, model perspective denoising, and denoising diffusion methods.
• Figure 2: Structure of DDRM. The left part is the backend recommender model. DDRM accepts both user and item embeddings as inputs and subsequently produces denoised embeddings that are fed back into the model to do the recommendation task.
• Figure 3: Performance comparison of noisy training with random noises in Yelp.
• Figure 4: Contributions of reconstruction loss and reweighted loss to DDRM compared with backend models.
• Figure 5: Contributions of user and item denoising modules to DDRM compared with backend models.