Heriot-Watt University, UK
Title: Proximal optimal transport divergences
Abstract: We introduce proximal optimal transport divergence, a novel discrepancy measure that interpolates between information divergences and optimal transport distances via an infimal convolution formulation. This divergence provides a principled foundation for optimal transport proximals and proximal optimization methods frequently used in generative modeling. We explore its mathematical properties and computational tractability, and establish connections to primal-dual formulation and adversarial learning. Building on the Benamou-Brenier dynamic formulation of optimal transport cost, we also establish a dynamic formulation for proximal OT divergences. Our framework generalizes existing approaches while offering new insights and computational tools for generative modelling and gradient-based learning in probability spaces.
TU Graz, AT
Title: Sampling versus optimization
Abstract: Motivated by the problem of solving Bayesian formulations of inverse problems in imaging, this talk presents several novel sampling algorithms inspired by advances in optimization. The first approach, called the "Gaussian Latent Machine", draws on concepts from
half-quadratic optimization - a powerful technique for transforming non-smooth or non-convex problems into more tractable subproblems. The second method, coined the "Inertial Langevin Algorithm", integrates ideas from inertial and accelerated gradient methods to significantly enhance the convergence speed of the widely used Langevin sampling framework.
Title: Physical models and machine learning for photography and astronomy
Abstract: We live in an era of data-driven approaches to image analysis, where modeling is sometimes considered obsolete. I will propose in this talk giving back to accurate physical models of image formation their rightful place next to machine learning in the overall processing and interpretation pipeline, and discuss two applications: super-resolution and high-dynamic range imaging from raw photographic bursts, and exoplanet detection and characterization in direct imaging at high contrast.
University of Eastern Finland, FI
Title: Utilising Monte Carlo method for light transport in optical imaging
Abstract: We study the inverse problems in two optical imaging modalities: diffuse optical tomography and quantitative photoacoustic tomography, when the forward operator is the Monte Carlo method for light transport. Monte Carlo a stochastic method that can be used to simulate the solution of the radiative transfer equation. In the approach, paths of photons are simulated when they undergo absorption and scattering events in a scattering medium. In the inverse problems optical tomography and quantitative photoacoustic tomography, absorption and scattering distributions inside the target medium are estimated. Now, due to the stochastic nature of the forward operator, also the search direction of a minimisation algorithm for solving these estimates is stochastic. The approach is studied with numerical simulations.