Contents
Title Page
Abstract
Introduction
I. What is a Neural Network?
II. Training Neural Networks to Solve Problems
III. The Basics of Evolution Strategies
Preliminary Research
I. Existing Visualization Systems
II. Motivations
III. Prototypes
Program Design
I. Class & Module Design
II. User Interface Design
Program Implementation
I. Porting
II. Implementation of
TESGraphWindow
III. Implementation of
TNVisWindow
IV. Implementation of
THeredityWindow
&
TGenerationWindow
V. Implementation of
TFamilyTreeWindow
VI. Implementation of
TESWindow
Results
I. Problems Solved
II. Observations
A. A Voyage into Search Space
B. Understanding Genetic Drift
C. Changing Weights & Local Optima
III. The power of
NVIS
A. Topological Optimization
B. Designing Networks
C. Extracting Domain Knowledge
Future Work
I. Known Bugs
II. Known Limitations
III. Possible Optimizations
IV. Potential Features
Conclusions
Bibliography
Appendix A: Timelines
I. Development Timeline
I. Experiment Timeline
Appendix B: Program source code (online version only)