“Graph Spectral Diffusion Model Provides High-Quality Data Generation with Lower Computational Consumption”

https://pubmed.ncbi.nlm.nih.gov/38117624

The Graph Spectral Diffusion Model (GSDM) proposed by Luo et al. generates high-quality graph-structured data by using low-rank diffusion SDEs on the graph spectrum space, providing a state-of-the-art performance with less computational consumption compared to existing models.