Interpolating Neural Network Construction Independent of Dataset Size
Publication date
Authors
DOI
Document Type
Bachelor Thesis
Metadata
Show full item recordCollections
License
CC-BY-NC-ND
Abstract
The thesis introduces an algorithm that constructs a three-layer neural network in polynomial time. It is proven that under the threshold activation function the size of the constructed neural network depends only on the geometric relationship between the two classes, rather than the size of the dataset.