Date of Award

8-1983

Degree Name

Doctor of Philosophy

Department

Mathematics

First Advisor

Dr. Dionysios Kountanis

Second Advisor

Dr. Mike Stoline

Third Advisor

Dr. Kung-Wei Yang

Fourth Advisor

Dr. Alden Wright

Abstract

Many problems in adaptive control, pattern recognition, filtering, identification, and artificial intelligence can be viewed as parameter optimization problems. The learning automation approach to these problems has two distinct advantages over the classic hill climbing methods: (1) the parameter space need not be metric and (2) a global rather than local optimum can be found. Unfortunately, these advantages do not come without corresponding difficulties, namely the problem of high dimensionality. A hierarchical system of learning automata has been used to reduce this problem somewhat, but inefficiencies still remain since the current hierarchical system was balanced and static hierarchical learning automaton structures. To resolve this problem, hierarchical learning automata which employ unbalanced and dynamic structures are introduced. Simulation results are also provided which show a significant reduction in convergence time when these hierarchical automata strategies are used.

Comments

Fifth Advisor: Dr. Ken Williams

Sixth Advisor: Dr. Carl Page

Access Setting

Dissertation-Open Access

Download

Share

COinS