A hybrid model approach to generalization in sequence learning

Both recurrent neural networks and humans are able to learn sequential information and generalize to sequences they have not experienced in training. However, they sometimes seem to differ in the way they perform generalization. A new hybrid model is introduced that relies on both a recurrent neural...

Full description

Saved in:
Bibliographic Details
Published inIJCNN'01. International Joint Conference on Neural Networks. Proceedings (Cat. No.01CH37222) Vol. 4; pp. 2393 - 2398 vol.4
Main Authors Spiegel, R., McLaren, I.P.L.
Format Conference Proceeding
LanguageEnglish
Published IEEE 2001
Subjects
Artificial intelligence
Backpropagation algorithms
Context modeling
Humans
Intelligent networks
Neural networks
Predictive models
Psychology
Recurrent neural networks
Testing
Online AccessGet full text

Cover

More Information
Summary:Both recurrent neural networks and humans are able to learn sequential information and generalize to sequences they have not experienced in training. However, they sometimes seem to differ in the way they perform generalization. A new hybrid model is introduced that relies on both a recurrent neural network and rules typically applied by human subjects.
ISBN:0780370449
9780780370449
ISSN:1098-7576
DOI:10.1109/IJCNN.2001.938741