Tree-Based Diffusion Schrödinger Bridge with Applications to Wasserstein Barycenters

Watch a technical seminar presentation where Mr Maxence Noble-Bourillot from Ecole Polytechnique Paris explores Tree-Based Diffusion Schrödinger Bridge (TreeDSB) and its applications to Wasserstein Barycenters. Dive into an advanced discussion of Multi-marginal Optimal Transport (mOT) and learn how TreeDSB extends the Diffusion Schrödinger Bridge algorithm to handle tree-structured quadratic costs. Discover how this methodology serves as a continuous state-space equivalent to the multimarginal Sinkhorn algorithm and its practical applications in high-dimensional scenarios like image interpolation and Bayesian fusion. The presentation, delivered at the INI Seminar as part of the data-driven engineering series, features collaborative work with researchers from ENS Ulm, Oxford University, and Ecole polytechnique.