Search:
Match:
1 results

Analysis

This paper addresses the cold-start problem in federated recommendation systems, a crucial challenge where new items lack interaction data. The proposed MDiffFR method leverages a diffusion model to generate embeddings for these items, guided by modality features. This approach aims to improve performance and privacy compared to existing methods. The use of diffusion models is a novel approach to this problem.
Reference

MDiffFR employs a tailored diffusion model on the server to generate embeddings for new items, which are then distributed to clients for cold-start inference.