Uncertainty-aware Guided Diffusion for Missing Data in Sequential Recommendation
Keywords: Diffusion Models, Recommender Systems, Missing Data
TL;DR: We introduce a dual-side Thompson sampling strategy to create uncertainty-aware guidance for DDMs in sequential recommendation.
Abstract: Denoising diffusion models (DDMs) have shown significant potential in generating oracle items that best match user preference with guidance from user historical interaction sequences. However, the quality of guidance is often compromised by the unpredictable missing data in the observed sequence, leading to suboptimal item generation. To tackle this challenge, we propose a novel uncertainty-aware guided diffusion model (DreamMiss) to alleviate the influence of missing data. The core of DreamMiss is the utilization of a dual-side Thompson sampling (DTS) strategy, which simulates the stochastical mechanism of missing data without disrupting preference evolution. Specifically, we first define dual-side probability models to capture user preference evolution, taking into account both local item continuity and global sequence stability. We then strategically remove items based on these two models with DTS, creating uncertainty-aware guidance for DDMs to generate oracle items. This can achieve DDMs’ consistency regularization, enabling them to resile against uncertain missing data. Additionally, to accelerate sampling in the reverse process, DreamMiss is implemented under the framework of denoising diffusion implicit models (DDIM). Extensive experimental results show that DreamMiss significantly outperforms baselines in sequential recommendation.
Primary Area: generative models
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 7669
Loading