Universität Karlsruhe
Projekt Fynesse

Selbständiges Lernen von Regelstrategien.

FYNESSE (FuzzY-NEuro-SyStEm) is an architecture that autonomously learns to make optimal sequential decisions in large domains. This problem of decision making is often encountered, e.g. control of dynamical systems or job shop scheduling. In contrary to supervised learning where an expert is needed to obtain desired input-output patterns for the decision system our reinforcement learning approach derives the control strategy in direct interaction with the process from success and failure.

Projektbeteiligte

Alumni
Dr. Martin Spott
Partner
Dr. Martin Riedmiller
Ralf Schoknecht

Publikationen zum Projekt

2000
Riedmiller, Spott, Weisbrod, Fynesse: A hybrid architecture for selflearning control
Login
Links