EDGE-Rec: Efficient and Data-Guided Edge Diffusion For Recommender Systems Graphs
Priyam, Utkarsh, Shah, Hemit, Botta, Edoardo
–arXiv.org Artificial Intelligence
Most recommender systems research focuses on binary historical user-item interaction encodings to predict future interactions. User features, item features, and interaction strengths remain largely under-utilized in this space or only indirectly utilized, despite proving largely effective in large-scale production recommendation systems. We propose a new attention mechanism, loosely based on the principles of collaborative filtering, called Row-Column Separable Attention RCSA to take advantage of real-valued interaction weights as well as user and item features directly. Building on this mechanism, we additionally propose a novel Graph Diffusion Transformer GDiT architecture which is trained to iteratively denoise the weighted interaction matrix of the user-item interaction graph directly. The weighted interaction matrix is built from the bipartite structure of the user-item interaction graph and corresponding edge weights derived from user-item rating interactions. Inspired by the recent progress in text-conditioned image generation, our method directly produces user-item rating predictions on the same scale as the original ratings by conditioning the denoising process on user and item features with a principled approach.
arXiv.org Artificial Intelligence
Sep-22-2024
- Country:
- North America > United States
- Pennsylvania > Allegheny County
- Pittsburgh (0.14)
- New York > New York County
- New York City (0.04)
- Pennsylvania > Allegheny County
- Europe
- Spain > Catalonia
- Barcelona Province > Barcelona (0.05)
- Germany > Bavaria
- Upper Bavaria > Munich (0.04)
- Spain > Catalonia
- Asia > China
- Jiangsu Province > Yancheng (0.04)
- North America > United States
- Genre:
- Workflow (0.47)
- Research Report (0.40)
- Industry:
- Media > Film (0.94)
- Leisure & Entertainment (0.69)
- Technology: