Nonlinear approximation and (deep) ReLU networks
From MaRDI portal
Publication:2117331
DOI10.1007/S00365-021-09548-ZzbMATH Open1501.41003arXiv1905.02199OpenAlexW3160447895MaRDI QIDQ2117331FDOQ2117331
Simon Foucart, Ronald DeVore, Ingrid Daubechies, B. Hanin, Guergana Petrova
Publication date: 21 March 2022
Published in: Constructive Approximation (Search for Journal in Brave)
Abstract: This article is concerned with the approximation and expressive powers of deep neural networks. This is an active research area currently producing many interesting papers. The results most commonly found in the literature prove that neural networks approximate functions with classical smoothness to the same accuracy as classical linear methods of approximation, e.g. approximation by polynomials or by piecewise polynomials on prescribed partitions. However, approximation by neural networks depending on n parameters is a form of nonlinear approximation and as such should be compared with other nonlinear methods such as variable knot splines or n-term approximation from dictionaries. The performance of neural networks in targeted applications such as machine learning indicate that they actually possess even greater approximation power than these traditional methods of nonlinear approximation. The main results of this article prove that this is indeed the case. This is done by exhibiting large classes of functions which can be efficiently captured by neural networks where classical nonlinear methods fall short of the task. The present article purposefully limits itself to studying the approximation of univariate functions by ReLU networks. Many generalizations to functions of several variables and other activation functions can be envisioned. However, even in this simplest of settings considered here, a theory that completely quantifies the approximation power of neural networks is still lacking.
Full work available at URL: https://arxiv.org/abs/1905.02199
Recommendations
- Error bounds for approximations with deep ReLU networks
- A note on the expressive power of deep rectified linear unit networks in high-dimensional spaces
- Approximation spaces of deep neural networks
- Provable approximation properties for deep neural networks
- Deep ReLU networks and high-order finite element methods
Artificial neural networks and deep learning (68T07) Neural networks for/in biological studies, artificial life and related topics (92B20) Approximation by other special function classes (41A30) Rate of convergence, degree of approximation (41A25) Approximation by arbitrary nonlinear expressions; widths and entropy (41A46) Neural nets applied to problems in time-dependent statistical mechanics (82C32)
Cites Work
- Title not available (Why is that?)
- Title not available (Why is that?)
- Multilayer feedforward networks are universal approximators
- Approximation by superpositions of a sigmoidal function
- Title not available (Why is that?)
- Optimal nonlinear approximation
- The Takagi function: a survey
- Weierstrass' function and chaos
- Wavelet compression and nonlinear \(n\)-widths
- Title not available (Why is that?)
- Error bounds for approximations with deep ReLU networks
- Neural Networks for Localized Approximation
- Deep learning in high dimension: Neural network expression rates for generalized polynomial chaos expansions in UQ
- Optimal Approximation with Sparsely Connected Deep Neural Networks
- Deep vs. shallow networks: an approximation theory perspective
- Provable approximation properties for deep neural networks
- Exponential convergence of the deep neural network approximation for analytic functions
- Deep Network Approximation for Smooth Functions
- Deep Network Approximation Characterized by Number of Neurons
Cited In (63)
- A multivariate Riesz basis of ReLU neural networks
- Provable Training of a ReLU Gate with an Iterative Non-Gradient Algorithm
- Collocation approximation by deep neural ReLU networks for parametric and stochastic PDEs with lognormal inputs
- Deep ReLU networks and high-order finite element methods. II: Chebyšev emulation
- Alternating minimization for regression with tropical rational functions
- Robust nonparametric regression based on deep ReLU neural networks
- On the latent dimension of deep autoencoders for reduced order modeling of PDEs parametrized by random fields
- Estimating a regression function in exponential families by model selection
- Improving the expressive power of deep neural networks through integral activation transform
- Approximation results for gradient flow trained shallow neural networks in \(1d\)
- Low dimensional approximation and generalization of multivariate functions on smooth manifolds using deep ReLU neural networks
- Constructive deep ReLU neural network approximation
- Error bounds for approximations using multichannel deep convolutional neural networks with downsampling
- Random neural networks in the infinite width limit as Gaussian processes
- Stable recovery of entangled weights: towards robust identification of deep neural networks from minimal samples
- On the SQH method for solving optimal control problems with non-smooth state cost functionals or constraints
- Exponential ReLU DNN expression of holomorphic maps in high dimension
- A functional equation with polynomial solutions and application to neural networks
- A note on the applications of one primary function in deep neural networks
- Deep Neural Network Surrogates for Nonsmooth Quantities of Interest in Shape Uncertainty Quantification
- Title not available (Why is that?)
- Deep Neural Network Approximation Theory
- Approximation spaces of deep neural networks
- A convergent deep learning algorithm for approximation of polynomials
- Theoretical issues in deep networks
- Spline representation and redundancies of one-dimensional ReLU neural network models
- Deep ReLU Networks Overcome the Curse of Dimensionality for Generalized Bandlimited Functions
- Mesh-informed neural networks for operator learning in finite element spaces
- Expressivity of Deep Neural Networks
- Approximation capabilities of neural networks on unbounded domains
- Machine learning design of volume of fluid schemes for compressible flows
- Universal approximation with quadratic deep networks
- Deep ReLU neural network approximation in Bochner spaces and applications to parametric PDEs
- Convergence of deep convolutional neural networks
- Approximation error for neural network operators by an averaged modulus of smoothness
- Deep learning via dynamical systems: an approximation perspective
- Deep learning-based approximation of Goldbach partition function
- ReLU deep neural networks from the hierarchical basis perspective
- Title not available (Why is that?)
- Information theory and recovery algorithms for data fusion in Earth observation
- A deep learning approach to Reduced Order Modelling of parameter dependent partial differential equations
- Neural network approximation
- Connections between numerical algorithms for PDEs and neural networks
- High-dimensional distribution generation through deep neural networks
- Approximation of compositional functions with ReLU neural networks
- ReLU neural networks of polynomial size for exact maximum flow computation
- Sparse Deep Neural Network for Nonlinear Partial Differential Equations
- Best \(n\)-term approximation of diagonal operators and application to function spaces with mixed smoothness
- Universality of gradient descent neural network training
- Title not available (Why is that?)
- Simultaneous neural network approximation for smooth functions
- Neural Parametric Fokker--Planck Equation
- Approximation by Combinations of ReLU and Squared ReLU Ridge Functions With <inline-formula> <tex-math notation="LaTeX">$\ell^1$ </tex-math> </inline-formula> and <inline-formula> <tex-math notation="LaTeX">$\ell^0$ </tex-math> </inline-formula> Controls
- Designing rotationally invariant neural networks from PDEs and variational methods
- Neural ODE Control for Classification, Approximation, and Transport
- Exponential ReLU neural network approximation rates for point and edge singularities
- Thermodynamically consistent physics-informed neural networks for hyperbolic systems
- A mesh-free method using piecewise deep neural network for elliptic interface problems
- Transferable neural networks for partial differential equations
- Approximation in shift-invariant spaces with deep ReLU neural networks
- Linearized two-layers neural networks in high dimension
- Optimal stable nonlinear approximation
- The construction and approximation of ReLU neural network operators
Uses Software
This page was built for publication: Nonlinear approximation and (deep) ReLU networks
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2117331)
