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

Domains of holomorphy (32D05) Approximation by polynomials (41A10) Rate of convergence, degree of approximation (41A25) Approximation by arbitrary nonlinear expressions; widths and entropy (41A46) Optimality conditions for problems involving randomness (49K45) Inverse problems in optimal control (49N45) Optimal transportation (49Q22) Sampling theory, sample surveys (62D05) Algorithms for approximation of functions (65D15)

arXiv classification

cs.NA math.NA math.ST stat.TH

Abstract: For two probability measures

h o

and

p i

with analytic densities on the

d

-dimensional cube

[- 1, 1]^{d}

, we investigate the approximation of the unique triangular monotone Knothe-Rosenblatt transport

T : [- 1, 1]^{d} o [- 1, 1]^{d}

, such that the pushforward

T_{s} h a r p h o

equals

p i

. It is shown that for

d i n m a t h b b N

there exist approximations

i l d e T

of

T

, based on either sparse polynomial expansions or deep ReLU neural networks, such that the distance between

i l d e T_{s} h a r p h o

and

p i

decreases exponentially. More precisely, we prove error bounds of the type

(or

for neural networks), where

N

refers to the dimension of the ansatz space (or the size of the network) containing

i l d e T

; the notion of distance comprises the Hellinger distance, the total variation distance, the Wasserstein distance and the Kullback-Leibler divergence. Our construction guarantees

i l d e T

to be a monotone triangular bijective transport on the hypercube

[- 1, 1]^{d}

. Analogous results hold for the inverse transport

S = 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.

This page was built for publication: Sparse approximation of triangular transports. Part I: the finite dimensional case

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6342739)