Statements (16)
| Predicate | Object |
|---|---|
| gptkbp:instanceOf |
gptkb:mathematical_concept
|
| gptkbp:activationFunctionRequirement |
nonconstant, bounded, and continuous
|
| gptkbp:alsoKnownAs |
gptkb:Universal_Approximation_Theorem
|
| gptkbp:author |
gptkb:Kurt_Hornik
|
| gptkbp:citation |
thousands of research papers
|
| gptkbp:field |
gptkb:machine_learning
neural networks |
| https://www.w3.org/2000/01/rdf-schema#label |
Hornik's theorem
|
| gptkbp:impact |
established theoretical foundation for neural networks as universal function approximators
|
| gptkbp:publicationYear |
1989
|
| gptkbp:publishedIn |
gptkb:Neural_Networks
|
| gptkbp:relatedTo |
gptkb:Universal_Approximation_Theorem
Cybenko's theorem |
| gptkbp:sentence |
A feedforward neural network with a single hidden layer containing a finite number of neurons can approximate any continuous function on compact subsets of R^n, under mild assumptions on the activation function.
|
| gptkbp:bfsParent |
gptkb:Cybenko_theorem
|
| gptkbp:bfsLayer |
7
|