Multilayer feedforward networks are universal approximators

Neural Networks

Volumn 2, Issue 5, 1989, Pages 359-366

  Hornik, Kurt ^a   Stinchcombe, Maxwell ^b   White, Halbert ^b  

^a VIENNA UNIVERSITY OF TECHNOLOGY   (Austria)

^b UNIVERSITY OF CALIFORNIA   (United States)

Author keywords

Back propagation networks; Feedforward networks; Mapping networks; Network representation capability; Sigma Pi networks; Squashing functions; Stone Weierstrass Theorem; Universal approximation

Indexed keywords

MATHEMATICAL TECHNIQUES--APPROXIMATION THEORY; MATHEMATICAL TRANSFORMATIONS;

BACK-PROPAGATION NETWORKS; FEEDFORWARD NETWORKS; MAPPING NETWORKS; SQUASHING FUNCTIONS, SIGMA-PI NETWORKS; STONE-WEIERSTRASS THEOREM; UNIVERSAL APPROXIMATORS;

SYSTEMS SCIENCE AND CYBERNETICS;

EID: 0024880831     PISSN: 08936080     EISSN: None     Source Type: Journal    
DOI: 10.1016/0893-6080(89)90020-8     Document Type: Article

References (26)

1. Approximation by superpositions of a sigmoidal function
2. There exists a neural network that does not make avoidable mistables
3. Kolmogorov's mapping neural network existence theorem
4. Theory of the back propagation neural network
5. Multilayer feedforward networks are universal approximators
6. Capabilities of three layer perceptrons
7. On the representation of continuous functions of many variables by superposition of continuous functions of one variable and addition
8. ϵ-entropy and ϵ-capacity of sets in functional spaces
9. How neural networks work
10. Modeles connexionistes de l'apprentissage
11. The thirteenth problem of Hilbert
12. Nonlinear dynamics of artificial neural systems
13. Convergence rates of maximum likelihood and related estimates in general parameter spaces
14. Universal approximation using feedforward networks with non-sigmoid hidden layer activation functions
15. The case for conceptual and operational separation of network architectures and learning mechanisms
16. Multilayer feedforward networks can learn arbitrary mappings: Connectionist nonparametric regression with automatic and semi-automatic determination of network complexity
17. The logic of activation functions