Towards A Tri-View Diffusion Framework for Recommendation

TL;DR

TV-Diff addresses diffusion-based recommender systems by introducing a tri-view framework that unifies thermodynamic optimization, topology-aware denoising, and principled hard-negative sampling. It defines Helmholtz free energy $H = U(R') - t S(R')$, employs an anisotropic denoiser to respect bipartite graph topology, and uses an Acceptance-Rejection Gumbel Sampling Process for informative negatives, all within a diffusion modeling setting. Empirical results on five real datasets show TV-Diff consistently outperforms baselines in accuracy and efficiency, with ablations validating each component. By reconciling energy and entropy while leveraging graph topology and targeted negatives, TV-Diff advances diffusion-based recommendations toward stronger generalization and practicality.

Abstract

Diffusion models (DMs) have recently gained significant interest for their exceptional potential in recommendation tasks. This stems primarily from their prominent capability in distilling, modeling, and generating comprehensive user preferences. However, previous work fails to examine DMs in recommendation tasks through a rigorous lens. In this paper, we first experimentally investigate the completeness of recommender models from a thermodynamic view. We reveal that existing DM-based recommender models operate by maximizing the energy, while classic recommender models operate by reducing the entropy. Based on this finding, we propose a minimalistic diffusion framework that incorporates both factors via the maximization of Helmholtz free energy. Meanwhile, to foster the optimization, our reverse process is armed with a well-designed denoiser to maintain the inherent anisotropy, which measures the user-item cross-correlation in the context of bipartite graphs. Finally, we adopt an Acceptance-Rejection Gumbel Sampling Process (AR-GSP) to prioritize the far-outnumbered unobserved interactions for model robustness. AR-GSP integrates an acceptance-rejection sampling to ensure high-quality hard negative samples for general recommendation tasks, and a timestep-dependent Gumbel Softmax to handle an adaptive sampling strategy for diffusion models. Theoretical analyses and extensive experiments demonstrate that our proposed framework has distinct superiority over baselines in terms of accuracy and efficiency.

Figures (11)

• Figure 1: Toy illustration of 2D visualization on items.
• Figure 2: Visualization of Energy (upper) and Entropy (lower) on the user's original and reconstructed interaction probability.
• Figure 3: Detailed comparison among different models. $\Delta U_* \uparrow \in \mathbb{R}^{-}$ or $\Delta S_* \downarrow \in \mathbb{R}^{+}$ denotes the energy or entropy in difference on training, i.e., $U_*(\hat{\textbf{R}'})-U_*(\hat{\textbf{R}})$ or $S_*(\hat{\textbf{R}'})-S_*(\hat{\textbf{R}})$. The element in blocks denotes the $\frac{y}{x}$.
• Figure 4: Ablation study on different views.
• Figure 5: Ablation study on entropy-based objectives.

Theorems & Definitions (12)

• lemma 1
• theorem 1
• Remark
• theorem 2
• theorem 3
• Remark
• proposition 1: graph_spectral_reference
• proposition 2
• theorem 4
• Remark