Score-based diffusion models are reshaping physics by transforming random noise into realistic simulations. Unlike other deep generative models, which often struggle with complex data distributions, diffusion models generate data through a gradual denoising process governed by stochastic differential equations. In this talk, I will begin with a brief comparison of the major generative modelling approaches before focusing on the core principles of diffusion models. I will then highlight their growing impact through recent applications in physics, showing how they are opening new doors for scientific discovery.
Organized by: Catarina Cosme