Discrete Bayesian Sample Inference for Graph Generation
3 days ago
- #Graph Generation
- #Machine Learning
- #Bayesian Inference
- GraphBSI is a novel one-shot graph generative model based on Bayesian Sample Inference (BSI).
- It refines a belief over graphs in the continuous space of distribution parameters, handling discrete structures naturally.
- GraphBSI is formulated as a stochastic differential equation (SDE) with a noise-controlled family of SDEs preserving marginal distributions.
- Theoretical analysis connects GraphBSI to Bayesian Flow Networks and Diffusion models.
- Empirical evaluation shows state-of-the-art performance on molecular and synthetic graph generation, outperforming existing models on Moses and GuacaMol benchmarks.