arXiv subjects

Bertram Tschiderer

Publications and source records attributed to Bertram Tschiderer.

9 recordsLinked to original sources

arXiv2024

The decomposition of stretched Brownian motion into Bass martingales

In previous work J. Backhoff-Veraguas, M. Beiglb\"ock and the present authors showed that the notions of stretched Brownian motion and Bass martingale between two probability measures on Euclidean space coincide if and only if these two measures satisfy an irreducibility condition. Now we consider the general case, i.e. a pair of measures which are not necessarily irreducible. We show that there is a paving of Euclidean space into relatively open convex sets such that the stretched Brownian motion decomposes into a (possibly uncountable) family of Bass martingales on these sets. This paving coincides with the irreducible convex paving studied in previous work of H. De March and N. Touzi.

math.PR↗

arXiv2024

$q$-Bass martingales

An intriguing question in martingale optimal transport is to characterize the martingale with prescribed initial and terminal marginals whose transition kernel is as Gaussian as possible. In this work we address an extension of this question, in which the role of the Gaussian distribution is replaced by an arbitrary reference measure $q$. Our first main result is a dual formulation of the corresponding martingale optimization problem in terms of convex functions. In the well-studied case when $q$ is Gaussian, the careful analysis of the solution to the above-mentioned optimization problem is a crucial building block in the construction of Bass martingales, i.e., Brownian martingales induced by gradients of convex functions with possibly non-degenerate starting laws. In our second main result we extend this concept beyond the Gaussian case by introducing the notion of $q$-Bass martingales in discrete time, and give sufficient conditions for their existence.

math.PR↗

arXiv2023

Diffusion processes as Wasserstein gradient flows via stochastic control of the volatility matrix

We consider a class of time-homogeneous diffusion processes on $\mathbb{R}^{n}$ with common invariant measure but varying volatility matrices. In Euclidean space, we show via stochastic control of the diffusion coefficient that the corresponding flow of time-marginal distributions admits an entropic gradient flow formulation in the quadratic Wasserstein space if the volatility matrix of the diffusion is the identity. After equipping $\mathbb{R}^{n}$ with a Riemannian metric, we prove that the diffusion process can be viewed as a gradient flow in the inherited Wasserstein space if the volatility matrix is the inverse of the underlying metric tensor. In the Euclidean case, our probabilistic result corresponds to the gradient flow characterization of the Fokker$-$Planck equation, first discovered in a seminal paper by Jordan, Kinderlehrer, and Otto. In the Riemannian setting, the corresponding result on the level of partial differential equations was established by Lisini, building on the metric theory developed by Ambrosio, Gigli, and Savaré.

math.PR↗

arXiv2023

The Bass functional of martingale transport

An interesting question in the field of martingale optimal transport, is to determine the martingale with prescribed initial and terminal marginals which is most correlated to Brownian motion. Under a necessary and sufficient irreducibility condition, the answer to this question is given by a $\textit{Bass martingale}$. At an intuitive level, the latter can be imagined as an order-preserving and martingale-preserving space transformation of an underlying Brownian motion starting with an initial law $α$ which is tuned to ensure the marginal constraints. In this article we study how to determine the aforementioned initial condition $α$. This is done by a careful study of what we dub the $\textit{Bass functional}$. In our main result we show the equivalence between the existence of minimizers of the Bass functional and the existence of a Bass martingale with prescribed marginals. This complements the convex duality approach in a companion paper by the present authors together with M. Beiglböck, with a purely variational perspective. We also establish an infinitesimal version of this result, and furthermore prove the displacement convexity of the Bass functional along certain generalized geodesics in the $2$-Wasserstein space.

math.PR↗

arXiv2023

Existence of Bass martingales and the martingale Benamou$-$Brenier problem in $\mathbb{R}^{d}$

In classical optimal transport, the contributions of Benamou$-$Brenier and McCann regarding the time-dependent version of the problem are cornerstones of the field and form the basis for a variety of applications in other mathematical areas. In this article, we characterize solutions to the martingale Benamou$-$Brenier problem as $\textit{Bass martingales}$, i.e. transformations of Brownian motion through the gradient of a convex function. Our result is based on a new (static) Brenier-type theorem for a particular weak martingale optimal transport problem. As in the classical case, the structure of the primal optimizer is derived from its dual counterpart, whose derivation forms the technical core of this article. A key challenge is that dual attainment is a subtle issue in martingale optimal transport, where dual optimizers may fail to exist, even in highly regular settings.

math.PR↗

arXiv2021

A trajectorial approach to relative entropy dissipation of McKean$\boldsymbol{-}$Vlasov diffusions: gradient flows and HWBI inequalities

We formulate a trajectorial version of the relative entropy dissipation identity for McKean$-$Vlasov diffusions, extending the results of the papers [FJ16,KST20a], which apply to non-interacting diffusions. Our stochastic analysis approach is based on time-reversal of diffusions and Lions' differential calculus over Wasserstein space. It allows us to compute explicitly the rate of relative entropy dissipation along every trajectory of the underlying diffusion via the semimartingale decomposition of the corresponding relative entropy process. As a first application, we obtain a new interpretation of the gradient flow structure for the granular media equation, generalizing the formulation in [KST20a] developed for the linear Fokker$-$Planck equation. Secondly, we show how the trajectorial approach leads to a new derivation of the HWBI inequality, which relates relative entropy (H), Wasserstein distance (W), barycenter (B) and Fisher information (I).

math.PR↗

arXiv2020

A variational characterization of Langevin$\boldsymbol{-}$Smoluchowski diffusions

We show that Langevin$-$Smoluchowski measure on path space is invariant under time-reversal, followed by stochastic control of the drift with a novel entropic-type criterion. Repeated application of these forward-backward steps leads to a sequence of stochastic control problems, whose initial/terminal distributions converge to the Gibbs probability measure of the diffusion, and whose values decrease to zero along the relative entropy of the Langevin$-$Smoluchowski flow with respect to this Gibbs measure.

math.PR↗

arXiv2020

A trajectorial approach to the gradient flow properties of Langevin-Smoluchowski diffusions

We revisit the variational characterization of conservative diffusion as entropic gradient flow and provide for it a probabilistic interpretation based on stochastic calculus. It was shown by Jordan, Kinderlehrer, and Otto that, for diffusions of Langevin-Smoluchowski type, the Fokker-Planck probability density flow maximizes the rate of relative entropy dissipation, as measured by the distance traveled in the ambient space of probability measures with finite second moments, in terms of the quadratic Wasserstein metric. We obtain novel, stochastic-process versions of these features, valid along almost every trajectory of the diffusive motion in the backward direction of time, using a very direct perturbation analysis. By averaging our trajectorial results with respect to the underlying measure on path space, we establish the maximal rate of entropy dissipation along the Fokker-Planck flow and measure exactly the deviation from this maximum that corresponds to any given perturbation. As a bonus of our trajectorial approach we derive the HWI inequality relating relative entropy (H), Wasserstein distance (W) and relative Fisher information (I).

math.PR↗

arXiv2020

Trajectorial Otto calculus

We revisit the variational characterization of diffusion as entropic gradient flux and provide for it a probabilistic interpretation based on stochastic calculus. It was shown by Jordan, Kinderlehrer, and Otto that, for diffusions of Langevin-Smoluchowski type, the Fokker-Planck probability density flow minimizes the rate of relative entropy dissipation, as measured by the distance traveled in the ambient space of probability measures with finite second moments, in terms of the quadratic Wasserstein metric. We obtain novel, stochastic-process versions of these features, valid along almost every trajectory of the diffusive motion in both the forward and, most transparently, the backward, directions of time, using a very direct perturbation analysis. By averaging our trajectorial results with respect to the underlying measure on path space, we establish the minimum rate of entropy dissipation along the Fokker-Planck flow and measure exactly the deviation from this minimum that corresponds to any given perturbation. As a bonus of our perturbation analysis we derive the so-called HWI inequality relating relative entropy (H), Wasserstein distance (W) and relative Fisher information (I).

math.PR↗