Generative models for scientific data can often benefit from using domain-informed constraints to reduce their problem space. In our new #NeurIPS2023 paper, we simplify constrained (Riemannian) diffusion models to make them better, faster & easier to use. arxiv.org/abs/2307.05439
1
34
219
25K
99
Download Image
Diffusion models have recently been applied to a wide range of scientific domains. In many of them, we may be able to improve model performance by making use of important, domain-informed constraints to reduce the problem space or guarantee a desired behaviour.