Marcelo Pereyra Informed Deep Generative AI Models for Statistical Inverse Problems: Applications in Computational Sensing…

We present a new framework that leverages modern generative AI to solve high-dimensional statistical inverse problems. The method combines Langevin diffusion and Markov chain Monte Carlo (MCMC) to build neural network architectures that generate samples closely following the target posterior distribution. The networks are modular and interpretable, with components representing learned generative priors and data likelihoods derived from forward measurement models, allowing measurement and instrument parameters to be specified at inference time. To improve efficiency, we use adversarial model distillation, enablingaccurate sampling with as few as four MCMC steps even in problems exceeding one million dimensions. We demonstrate the approach on challenging image and video restoration tasks.

Marcelo Pereyra is a Professor in Statistics and UKRI ESPRC Open Research Fellow at the School of Mathematical and Computer Sciences of Heriot-Watt University & Maxwell Institute for Mathematical Sciences. He leads pioneering research advancing the statistical foundations of quantitative and scientific imaging, shaping how image data are used as rigorous quantitative evidence, and forging deep connections between statistical, variational, and machine learning approaches to imaging. His leadership and contributions have been recognised through multiple prestigious awards, most recently a five-year fulltime EPSRC Open Fellowship to drive the next generation of breakthroughs in statistical imaging sciences based on physics-informed generative articial intelligence.

The event will be held in EM182 at 15:00 on the 30th September.