DEEP RELU NETWORKS OVERCOME THE CURSE OF DIMENSIONALITY FOR GENERALIZED BANDLIMITED FUNCTIONS

被引：17

作者：

Montanelli, Hadrien ^{[1
]}

Yang, Haizhao ^{[2
]}

Du, Qiang ^{[3
]}

机构：

[1] Ecole Polytech, Ctr Math Appl, Palaiseau, France

[2] Purdue Univ, Dept Math, W Lafayette, IN 47907 USA

[3] Columbia Univ, Dept Appl Phys & Appl Math, New York, NY USA

来源：

JOURNAL OF COMPUTATIONAL MATHEMATICS | 2021年 / 39卷 / 06期

关键词：

Machine learning; Deep ReLU networks; Curse of dimensionality; Approxima-tion theory; Bandlimited functions; Chebyshev polynomials; ERROR-BOUNDS; OPTIMAL APPROXIMATION; SUPERPOSITION; SMOOTH;

D O I：

10.4208/jcm.2007-m2019-0239

中图分类号：

O29 [应用数学];

学科分类号：

070104 ;

摘要：

We prove a theorem concerning the approximation of generalized bandlimited multivariate functions by deep ReLU networks for which the curse of the dimensionality is overcome. Our theorem is based on a result by Maurey and on the ability of deep ReLU networks to approximate Chebyshev polynomials and analytic functions efficiently.

引用

页码：801 / 815

页数：15

共 50 条

[31] ON DEEP LEARNING AS A REMEDY FOR THE CURSE OF DIMENSIONALITY IN NONPARAMETRIC REGRESSION
Bauer, Benedikt
Kohler, Michael
ANNALS OF STATISTICS, 2019, 47 (04): : 2261 - 2285
[32] A Proof that Artificial Neural Networks Overcome the Curse of Dimensionality in the Numerical Approximation of Black-Scholes Partial Differential Equations
Grohs, Philipp
Hornung, Fabian
Jentzen, Arnulf
von Wurstemberger, Philippe
MEMOIRS OF THE AMERICAN MATHEMATICAL SOCIETY, 2023, 284 (1410) : 1 - 106
[33] Why and when can deep-but not shallow-networks avoid the curse of dimensionality: A review
Poggio T.
Mhaskar H.
Rosasco L.
Miranda B.
Liao Q.
International Journal of Automation and Computing, 2017, 14 (5) : 503 - 519
[34] Approximation of compositional functions with ReLU neural networks
Gong, Qi
Kang, Wei
Fahroo, Fariba
SYSTEMS & CONTROL LETTERS, 2023, 175
[35] Error bounds for approximations with deep ReLU networks
Yarotsky, Dmitry
NEURAL NETWORKS, 2017, 94 : 103 - 114
[36] Generalized kernels for reconstructing bandlimited functions and their Hilbert transform
Boche, H
Protzmann, M
1998 CONFERENCE ON PRECISION ELECTROMAGNETIC MEASUREMENTS DIGEST, 1998, : 426 - 427
[37] Approximation in LP(μ) with deep ReLU neural networks
Voigtlaender, Felix
Petersen, Philipp
2019 13TH INTERNATIONAL CONFERENCE ON SAMPLING THEORY AND APPLICATIONS (SAMPTA), 2019,
[38] The recovery of ridge functions on the hypercube suffers from the curse of dimensionality
Doerr, Benjamin
Mayer, Sebastian
JOURNAL OF COMPLEXITY, 2021, 63
[39] Global optimization of objective functions represented by ReLU networks
Christopher A. Strong
Haoze Wu
Aleksandar Zeljić
Kyle D. Julian
Guy Katz
Clark Barrett
Mykel J. Kochenderfer
Machine Learning, 2023, 112 : 3685 - 3712
[40] Approximation By Bandlimited Functions, Generalized K-Functionals and Generalized Moduli of Smoothness
S. Artamonov
K. Runovski
H.-J. Schmeisser
Analysis Mathematica, 2019, 45 : 1 - 24

← 1 2 3 4 5 →