Collective Intelligence from a Population of Evolving Neural Networks

摘要

In the paper an agent-based system of evolving neural networks dedicated to time-series prediction is presented. First the idea of a multi-agent predicting system is introduced and selected methods of management of collective intelligence of agent population(s) are considered. Then the problems and possibilities of evolutionary design of a predicting neural network are shortly discussed. Finally a hybrid solution combining both approaches in an evolutionary multi-agent system (EMAS) is proposed. General considerations are illustrated by the particular realisation of such a system and selected experimental results conclude the work.