Evolving DNA Classifiers with Extinction Based Ring Optimization

Daniel Ashlock, Sierra Gillis, Jennifer Garner, and Gary Fogle
Submitted to CIBCB 2015

Abstract PDF eprint

Extinction is a natural process that drives biological evolution. In this study the impact of four different extinction operators, two novel for ring optimization, on the evolution of side effect machines with a ring optimizer were studied. Side effect machines are an emerging technology used to generate features for DNA classification. Ring optimization is a type of evolutionary algorithm inspired by the biological concept of ring species. Earlier work shows that ring optimization is an efficient technique for locating good side effect machines with substantial robustness against parameter choice for the optimizer. This study extends that research by incorporating extinction which has been shown to substantially improve the performance of the ring optimizer on discrete and numerical test problems. Two of four extinction operators are found to be able to improve the quality of the best outcome while all four are able to reset the ring optimizer into a more exploratory state.