Lorenz Richter speaks at RICAM workshop on generative AI
Our founder Lorenz Richter spoke at the workshop on generative AI held from 19 to 23 October 2026 at RICAM in Linz. RICAM is the Johann Radon Institute for Computational and Applied Mathematics of the Austrian Academy of Sciences (ÖAW). His talk was titled “A non-Markovian approach to diffusion-based sampling”.
Diffusion and flow models in scientific computing
The workshop is part of RICAM’s special semester “AI in Scientific Computing”. It focused on diffusion and flow models, the class of generative models behind many current image and molecule generators. These models learn to turn random noise step by step into realistic data. The programme covered their mathematical foundations, sampling and fine-tuning, uncertainty quantification and inverse problems, along with applications in biology and drug discovery.
dida’s contribution
In his talk, Lorenz presented a non-Markovian approach to diffusion-based sampling. Classical diffusion processes are Markovian: each step depends only on the current state. Non-Markovian methods drop this assumption. Other speakers came from TU Graz, Imperial College London, Technion, Aalto University, Microsoft Research and other institutions.
Thanks
We thank the organisers for the invitation and for a week of strong exchange between mathematics and machine learning. We hope to have more reasons to travel to Austria in the future.