The purpose of this document is to report about the use of
reactive agents and hybrid methods (i.e. artificial neural
networks (ANN) and genetic algorithms (GA)) in Multi Agent Systems
(MAS) in order to solve global classification problems. First we
give an overview about the various technologies and ideas that
were developed and show how other researchers used them.

Then we try to use them ourselves in order to create agents
suitable for global classification tasks. We discuss various
methods for doing so and in the end we create an agent that has
conservative and explorative behavior. An experiment shows that
our way of creating such an agent works.