Kolmogorov's theorem and multilayer neural networks

Neural Networks

Volumn 5, Issue 3, 1992, Pages 501-506

  Kůrková, Věra ^a  

^a INSTITUTE OF MATHEMATICS   (Czech Republic)

Author keywords

Approximations of continuous functions; Estimates of number of hidden units; Feedforward neural networks; Modulus of continuity; Multilayer perceptron type networks; Sigmoidal activation function; Uniform approximation; Universal approximation capabilities

Indexed keywords

MATHEMATICAL TECHNIQUES - APPLICATIONS;

FEEDFORWARD NEURAL NETWORKS; KOLMOGOROV THEOREM; MODULUS OF CONTINUITY; MULTILAYER PERCEPTRON TYPR NETWORKS; SIGMOIDAL ACTIVATION FUNCTION;

NEURAL NETWORKS;

EID: 0026627415     PISSN: 08936080     EISSN: None     Source Type: Journal    
DOI: 10.1016/0893-6080(92)90012-8     Document Type: Article

References (16)

1. On functions of three variables
2. Construction of neural nets using the Radon transform
3. Approximation by superpositions of a single function
4. On the approximate realization of continuous mappings by neural networks
5. Representation properties of networks Kolmogorov's theorem is irrelevant
6. Kolmogorov's mapping neural network existence theorem
7. Theory of the back-propagation neural network
8. Multilayer feedforward networks are universal approximators
9. Approximation capabilities of multilayer feedforward networks
10. On the representations of continuous functions of many variables by superpositions of continuous functions of one variable and addition
11. 13th Hilbert's problem and neural networks
12. Kolmogorov's Theorem is relevant
13. On the structure of continuous functions of several variables
14. Universal approximation using feedforward networks with non-sigmoid hidden layer activation functions