AA5 – Variational Problems in Data-Driven Applications



Convolutional Brenier Generative Networks

Project Heads

Hanno Gottschalk, Gabriele Steidl

Project Members

Ségolène Martin

Project Duration

01.01.2024 − 31.12.2025

Located at

TU Berlin


Brenier’s theorem from optimal transport has been used to prove the existence of perfect generators in generative adversarial learning. Only recently, Brenier potentials were approximated by input convex neural networks. The aim of this project is the design, mathematical study and application of deep convolutional neural networks for approximating Brenier-like maps, resp. potentials in view of equivariance, resp. invariance properties.

External Website

Selected Publications