Distributed systems and self-organization
June Power · 1990
This paper examines distributed systems as complex systems which need to evolve in order to adapt to unpredictable events. As a consequence, system management must concern itself not only with changing resources, but also with the way information changes through its use. The central theme is that distributed systems are systems of relations between objects which are commensurate with language. Information exchanged between entities is a reflection of a coupling between their organisational domains. It is this shared aspect of communication that forms a basis for evolving knowledge in the system. The Learning Classifier System (LCS) is introduced as a paradigm for evolving knowledge within this framework. The LCS is a learning system that learns rules, called classifiers, to guide its interaction in some specified environment. It builds up its components from significant bit clusters, called schemata, which represent the attributes of the underlying system. The paper concludes by suggesting the LCS as a prototype agent which cooperates with local access mechanisms to adapt to changes in the environment.