On a constructive proof of Kolmogorov's superposition theorem
From MaRDI portal
(Redirected from Publication:843717)
In 1957 Kolmogorov showed that any continuous function \(f:[0,1]^n\to \mathbb{R}\) can be represented as a composition of addition and some functions of one variable. His proof was not constructive. Recently Kolmogorov's superposition theorem was applied in the study of neural networks. Motivated by these applications, Sprecher [Neural Netw. 9, No. 5, 765--772 (1996); Neural. Netw. 10, No. 3, 447--457 (1997)] proposed a numerical algorithm which gives a constructive proof of the theorem. \textit{M. Köppen} [Lect. Notes Comput. Sci. 2415, 474--479 (2002; Zbl 1013.68811)] observed that the Sprecher argumentation contains some gaps and suggested modifications, but without any proof. In this paper the authors, using the idea of Köppen, show that Sprecher's algorithm converges. It seems, it is the first correct constructive proof of Kolmogorov's superposition theorem.
Recommendations
- Approximative versions of Kolmogorov's superposition theorem, proved constructively
- A universal mapping for Kolmogorov's superposition theorem
- Kolmogorov's superposition theorem and sums of algebras
- Computational aspects of Kolmogorov's superposition theorem
- A superposition theorem of Kolmogorov type for bounded continuous functions
- A constructive proof of the Gelfand--Kolmogorov theorem
- On the Kolmogorov theorem
- Publication:3484155
- Kolmogorov theorem revisited
Cites work
- An improvement in the superposition theorem of Kolmogorov
- Approximative versions of Kolmogorov's superposition theorem, proved constructively
- Dimension of metric spaces and Hilbert’s problem 13
- scientific article; zbMATH DE number 3147701 (Why is no real title available?)
- scientific article; zbMATH DE number 3147888 (Why is no real title available?)
- scientific article; zbMATH DE number 1928659 (Why is no real title available?)
- scientific article; zbMATH DE number 804856 (Why is no real title available?)
- scientific article; zbMATH DE number 3246460 (Why is no real title available?)
- scientific article; zbMATH DE number 964891 (Why is no real title available?)
- scientific article; zbMATH DE number 3054965 (Why is no real title available?)
- Implications and applications of Kolmogorov's superposition theorem
- On the Structure of Continuous Functions of Several Variables
Cited in
(38)- A universal mapping for Kolmogorov's superposition theorem
- Computational aspects of Kolmogorov's superposition theorem
- Approximative versions of Kolmogorov's superposition theorem, proved constructively
- On interpolation properties of exterior functions in Kolmogorov superpositions
- Error bounds for deep ReLU networks using the Kolmogorov-Arnold superposition theorem
- Kolmogorov's superposition theorem and sums of algebras
- A three layer neural network can represent any multivariate function
- From Hilbert's 13th problem to the theory of neural networks: constructive aspects of Kolmogorov's superposition theorem
- A version of Kolmogorov-Arnold representation theorem for differentiable functions of several variables
- scientific article; zbMATH DE number 1928659 (Why is no real title available?)
- scientific article; zbMATH DE number 804856 (Why is no real title available?)
- Deep network with approximation error being reciprocal of width to power of square root of depth
- Some aspects of approximation and interpolation of functions artificial neural networks
- Spaces \(\ell\)-dominated by \(I\) or \(\mathbb{R}\)
- Superposition, reduction of multivariable problems, and approximation
- Hilbert 13: are there any genuine continuous multivariate real-valued functions?
- High-discrepancy sequences for high-dimensional numerical integration
- Neural network approximation: three hidden layers are enough
- A note on computing with Kolmogorov superpositions without iterations
- The Kolmogorov-Arnold representation theorem revisited
- ExSpliNet: An interpretable and expressive spline-based neural network
- Kolmogorov-Arnold-informed neural network: a physics-informed deep learning framework for solving forward and inverse problems based on Kolmogorov-Arnold networks
- Parameterized analytical solution of seepage equation for reservoir simulation using physics-informed Kolmogorov-Arnold network without labels
- On the expressivity of the ExSpliNet KAN model
- Construction of the Kolmogorov-Arnold networks using the Newton-Kaczmarz method
- Gene mutation estimations via mutual information and Ewens sampling based CNN & machine learning algorithms
- Scaled-cPIKANs: spatial variable and residual scaling in Chebyshev-based physics-informed Kolmogorov-Arnold networks
- A Kolmogorov high order deep neural network for high frequency partial differential equations in high dimensions
- Guidelines for building a realistic algorithmic trading market simulator for backtesting while incorporating market impact: agent-based strategies in neural network format, ecosystem dynamics \& detection
- HYDRA: symbolic feature engineering of overparameterized Eulerian hyperelasticity models for fast inference time
- DeepOKAN: deep operator network based on Kolmogorov Arnold networks for mechanics problems
- Neural tangent kernel analysis to probe convergence in physics-informed neural solvers: PIKANs vs. PINNs
- PG-KINN: A Physics-Informed Petrov-Galerkin Kolmogorov-Arnold Network for Solving Forward and Inverse PDEs
- Bridging heavy tails and artificial intelligence
- Approximation and error bounds for Kolmogorov-Arnold networks applications to dynamical system discovery via linear multistep methods
- Wave dynamics in thermoelastic layers: a machine learning approach
- Physics informed Kolmogorov-Arnold neural networks for dynamical analysis via efficient-KAN and WAV-KAN
- On computational algorithms for real-valued continuous functions of several variables
This page was built for publication: On a constructive proof of Kolmogorov's superposition theorem
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q843717)
