Multilayer feedforward networks are universal approximators
@article{Hornik1989MultilayerFN,
title={Multilayer feedforward networks are universal approximators},
author={K. Hornik and M. Stinchcombe and H. White},
journal={Neural Networks},
year={1989},
volume={2},
pages={359-366}
}Abstract This paper rigorously establishes that standard multilayer feedforward networks with as few as one hidden layer using arbitrary squashing functions are capable of approximating any Borel measurable function from one finite dimensional space to another to any desired degree of accuracy, provided sufficiently many hidden units are available. In this sense, multilayer feedforward networks are a class of universal approximators.
Supplemental Presentations
Topics from this paper
15,287 Citations
Approximation capabilities of multilayer feedforward networks
- Mathematics, Computer Science
- Neural Networks
- 1991
- 3,863
- PDF
Multilayer Feedforward Networks with a Non-Polynomial Activation Function Can Approximate Any Function
- Mathematics, Computer Science
- Neural Networks
- 1993
- 1,300
- PDF
Neural networks with a continuous squashing function in the output are universal approximators
- Mathematics, Computer Science
- Neural Networks
- 2000
- 66
- PDF
A constructive method for multivariate function approximation by multilayer perceptrons
- Mathematics, Medicine
- IEEE Trans. Neural Networks
- 1992
- 100
Constructive Approximation of Discontinuous Functions by Neural Networks
- Mathematics, Computer Science
- Neural Processing Letters
- 2007
- 19
Function approximation using a partition of the input space
- Mathematics
- [Proceedings 1992] IJCNN International Joint Conference on Neural Networks
- 1992
Generalization and Approximation Capabilities of Multilayer Networks
- Mathematics, Computer Science
- Neural Computation
- 1993
- 16
References
SHOWING 1-10 OF 24 REFERENCES
Universal approximation using feedforward networks with non-sigmoid hidden layer activation functions
- Mathematics
- International 1989 Joint Conference on Neural Networks
- 1989
- 269
- PDF
There exists a neural network that does not make avoidable mistakes
- Mathematics, Computer Science
- IEEE 1988 International Conference on Neural Networks
- 1988
- 194
Approximation by superpositions of a sigmoidal function
- Mathematics, Computer Science
- Math. Control. Signals Syst.
- 1989
- 4,036
- PDF
Parallel distributed processing: explorations in the microstructure of cognition, vol. 1: foundations
- Computer Science
- 1986
- 15,317
- PDF
Approximation by superpositions of a sigmoidaifunction
- Urbana
- 1988
Capabilities of three layer percep - trans
- IEEE Second International Conference on Neural Networks
- 1988
Multilayer feedforward networks are universal approximators (Discussion Paper 88-45)
- Multilayer feedforward networks are universal approximators (Discussion Paper 88-45)
- 1988