Full Professor, Oxford, University of Oxford
3 papers at NeurIPS 2025
We extend the Iterative Markovian Fitting procedure for Schrödinger Bridge computation to the tree-structured setting, which notably includes entropic Wasserstein barycentres as a special case.
We present the Rao-Blackwellised reparameterisation gradient estimator and show its performance gains on a suite of probabilistic models.
We show that diffusion models adapt to low-dimensional data geometry as a result of how approximations are formed at the level of the score function. We show that smoothing the score function naturally produces smoothing along the data manifold.
