English

Sparse approximation of triangular transports. Part I: the finite dimensional case

Numerical Analysis 2021-07-29 v2 Numerical Analysis Statistics Theory Statistics Theory

Abstract

For two probability measures ρ\rho and π\pi with analytic densities on the dd-dimensional cube [1,1]d[-1,1]^d, we investigate the approximation of the unique triangular monotone Knothe-Rosenblatt transport T:[1,1]d[1,1]dT:[-1,1]^d\to [-1,1]^d, such that the pushforward TρT_\sharp\rho equals π\pi. It is shown that for dNd\in\mathbb{N} there exist approximations T~\tilde T of TT, based on either sparse polynomial expansions or deep ReLU neural networks, such that the distance between T~ρ\tilde T_\sharp\rho and π\pi decreases exponentially. More precisely, we prove error bounds of the type exp(βN1/d)\exp(-\beta N^{1/d}) (or exp(βN1/(d+1))\exp(-\beta N^{1/(d+1)}) for neural networks), where NN refers to the dimension of the ansatz space (or the size of the network) containing T~\tilde T; the notion of distance comprises the Hellinger distance, the total variation distance, the Wasserstein distance and the Kullback-Leibler divergence. Our construction guarantees T~\tilde T to be a monotone triangular bijective transport on the hypercube [1,1]d[-1,1]^d. Analogous results hold for the inverse transport S=T1S=T^{-1}. The proofs are constructive, and we give an explicit a priori description of the ansatz space, which can be used for numerical implementations.

Keywords

Cite

@article{arxiv.2006.06994,
  title  = {Sparse approximation of triangular transports. Part I: the finite dimensional case},
  author = {Jakob Zech and Youssef Marzouk},
  journal= {arXiv preprint arXiv:2006.06994},
  year   = {2021}
}

Comments

The original manuscript arXiv:2006.06994v1 has been split into two parts; the present paper is the first part

R2 v1 2026-06-23T16:15:58.857Z