Development of an FPGA-Based Hardware Evaluation System for Use with GA-Designed Artificial Neural Networks

Author

Dennis Duncan Earl, University of Tennessee - Knoxville

Date of Award

8-2004

Degree Type

Dissertation

Degree Name

Doctor of Philosophy

Major

Electrical Engineering

Major Professor

Donald Wayne Bouldin

Committee Members

Dr. Mongi Abidi, Dr. Seong-Gon Kong, Dr. Bruce MacLennan

Abstract

The Hardware-Evolved Digital Artificial Neural Network (HEDANN) design platform is a circuit design platform built to evolve complex architecture ANN circuits using re-configurable hardware. By using genetic algorithms to evolve complex architecture ANN designs in field programmable gate arrays, this system is the first design system to evolve physical ANN circuits with unconstrained network architectures. With the HEDANN design system, the evolution of ANNs with recursive, non-layered, complex architectural connections is made possible. In addition, the HEDANN design system is capable of evolving device-independent circuit designs that can operate properly across a wide range of operating temperatures. This system is presented as a powerful new tool for researchers working to develop both artificially intelligent systems and complex evolvable hardware. To demonstrate the potential benefits of this unique design platform, the details of two trial experiments are presented and the results discussed.

Recommended Citation

Earl, Dennis Duncan, "Development of an FPGA-Based Hardware Evaluation System for Use with GA-Designed Artificial Neural Networks. " PhD diss., University of Tennessee, 2004.
https://trace.tennessee.edu/utk_graddiss/2171