Topic: Evolutionary connectionism

Speaker: Angelos Molfetas

Abstract:

This presentation provides a review of research done in the domain of 
evolutionary connectionism. This review will comprise of an explanation 
of evolutionary connectionism, a discussion of the current state of research 
in this field and a discussion of the types of research questions and 
research processes that characterise this field. The combination of evolutionary 
and neural computing has introduced a new phase in the development of 
intelligent systems. By employing evolutionary connectionism, systems are 
able to produce designs for adaptive systems. Genetic algorithms (GAs) can 
be applied to artificial neural networks (ANNs) to derive various neural 
network characteristics including number of node layers, number of nodes 
per layer, learning algorithms and appropriate biases. Evolutionary 
connectionism has been applied in many areas including speech recognition, 
computational finance and games. Evolutionary computation has been used to 
produce learning systems capable of competing within certain environmental
frameworks without the use of expert knowledge.