Each seminar session is divided into two scientific presentations of 40 minutes each: 30 minutes of talk and 10 minutes of questions. François Lanusse & Johannes Hertrich will host the April 2025 session!
Registration is free but compulsory, subject to availability. A buffet will be served at the end of the seminar.
Abstract
TBA
Abstract
In order to sample from an unnormalized probability density function, we propose to combine continuous normalizing flows (CNFs) with rejection-resampling steps based on importance weights. We relate the iterative training of CNFs with regularized velocity fields to a proximal mappings in the Wasserstein space. The alternation of local flow steps and non-local rejection-resampling steps allows to overcome local minima and mode collapse for multimodal distributions. The arising model can be trained iteratively, reduces the reverse Kulback-Leibler (KL) loss function in each step, allows to generate iid samples and moreover allows for evaluations of the generated underlying density. Numerical examples demonstrate the efficiency of our approach.