Self-improvement for the ADATE automatic programming system
Ronald E. Olson, Brock Wilcox · 2002
Automatic Design of Algorithms Through Evolution (ADATE) is a system for automatic programming based on the neutral theory of evolution. This work examines methods of self-improvement for the ADATE system. Experiments are performed testing two of the suggested methods, named SIGrewriting and SIG-reshaping. SIG-rewriting synthesizes transformations to explore neutral networks, while SIG-reshaping creates a predicate for accepting or rejecting nonneutral mutations. It is found that SIGrewriting has created no significant improvement for the ADATE system, while SIGreshaping does have improved results.