Abstract of FKI-189-94
Document-Name: FKI-189-94 (not available via ftp)
Title: Some Studies in Distributed Machine Learning and
Authors: Gerhard Weiss
Category: Technical Report (Forschungsberichte Künstliche Intelligenz)
Abstract: This article focusses on the intersection of distributed
machine learning and organizational design in the context of
multi-agent systems. A computational approach to distributed
reinforcement learning from experience and interaction is
described and analyzed. According to this approach, multiple
agents coordinate their actions by collectively learning
appropriate instantiations of one of the most fundamental
organizational structures, namely, hierarchically arranged
groups of compatible agents. Distributed learning encompasses
two interrelated processes: credit assignment (i.e., the
process of approximating the goal relevance of the agents'
and the groups' activities) and group development (i.e., the
process of creating new and dissolving old groups). The
approach is formalized in an algorithm called DFG.
Theoretical and experimental results are presented which
demonstrate the learning abilities of this approach.
The approach is discussed and research directions are
Keywords: Multi-agent systems, distributed machine learning,
Size: 20 pages
Copyright: The ``Forschungsberichte Künstliche Intelligenz''
series includes primarily preliminary publications,
specialized partial results, and supplementary
material. In the interest of a subsequent final
publication these reports should not be copied. All
rights and the responsability for the contents of the
report are with the authors, which would appreciate