Scalability of learning impact on complex parameters in recurrent neural networks

Authors:
Branko Šter;Andrej Dobnikar
Affiliations:
Faculty of Computer and Information Science, University of Ljubljana, Ljubljana, Slovenia;Faculty of Computer and Information Science, University of Ljubljana, Ljubljana, Slovenia
Venue:
ICANNGA'09 Proceedings of the 9th international conference on Adaptive and natural computing algorithms
Year:
2009

Citing 4
Cited 0

Learning automata: an introduction

Learning automata: an introduction
Handbook of Graphs and Networks: From the Genome to the Internet

Handbook of Graphs and Networks: From the Genome to the Internet
Structural Properties of Recurrent Neural Networks

Neural Processing Letters
Recurrent neural-network training by a learning automaton approach for trajectory learning and control system design

IEEE Transactions on Neural Networks

Quantified Score

Hi-index	0.00

Visualization

Abstract

The impact of problem extents and network sizes on learning in recurrent neural networks is analysed in terms of structural parameters of related graphs. In previous work the influence of learning on the changes of the typical parameters such as characteristic path length, clustering coefficient, degree distribution and entropy, was investigated. In the present work the focus is enlarged to the scaling problem of the learning paradigm. The results prove the scalability of learning procedures due to the retained dynamics of the parameters during learning with different problem extents and network sizes.