OS Stochastische Analysis: Computational methods for Adapted Transport via regularization

Time
Tuesday, 8. February 2022
17:00 - 18:30

Location
R512

Organizer

Speaker:
Gudmund Pammer (ETH Zurich)

Abstract: In 2013 Cuturi, in his seminal work “Lightspeed Computation of Optimal Transport”, rendered Optimal Transport computationally tractable via the resolution of the entropically regularized version. Adapted Transport (which includes Causal Optimal Transport and the Adapted Wasserstein distance) is a variant of optimal transport specifically designed for distributions on path spaces. The driving motivation behind this variant is to provide a notion of distance between models that is suitable for problems in mathematical finance.

We will elaborate on the numerical resolution of Adapted Transport via regularization: We show consistency of the regularization, propose a novel numerical scheme based on adapted variations of the Sinkhorn algorithm and prove (quantitative) convergence results.

The talk is based on joint work with Stephan Eckstein.