Detail publikace

Neural Network for Text Associations

ZBOŘIL, F.

Originální název

Neural Network for Text Associations

Anglický název

Neural Network for Text Associations

Jazyk

en

Originální abstrakt

The paper proposes a new neural network based on the principle of the Restricted Coulomb Energy (RCE) classifier. The network has three layers of neurons and it works as a heteroassociative memory. The short principles of the proposed neural network - topology, learning and retrieving - are described in the paper. Experiments with text lines associations that have been done with this neural network and with the Bidirectional Associative Memory (BAM) and a comparison of these results are described in the paper, too.

Anglický abstrakt

The paper proposes a new neural network based on the principle of the Restricted Coulomb Energy (RCE) classifier. The network has three layers of neurons and it works as a heteroassociative memory. The short principles of the proposed neural network - topology, learning and retrieving - are described in the paper. Experiments with text lines associations that have been done with this neural network and with the Bidirectional Associative Memory (BAM) and a comparison of these results are described in the paper, too.

Dokumenty

BibTex 


@inproceedings{BUT5418,
  author="František {Zbořil}",
  title="Neural Network for Text Associations",
  annote="The paper proposes a new neural network based on the principle of the Restricted Coulomb Energy (RCE) classifier. The network has three layers of neurons and it works as a heteroassociative memory. The short principles of the proposed neural network - topology, learning and retrieving - are described in the paper. Experiments with text lines associations that have been done with this neural network and with the Bidirectional Associative Memory (BAM) and a comparison of these results are described in the paper, too.",
  booktitle="Proceedings of XXIInd International Colloquium ASIS 2000",
  chapter="5418",
  year="2000",
  month="january",
  pages="145--150",
  type="conference paper"
}