Matching Hyper-heuristics and Genetic Programming

Presenter Information

Travis Bueter

Department

Computer Science

Major

Computer Science

Research Advisor

Tauritz, Daniel R.

Advisor's Department

Computer Science

Abstract

Modern society is faced with ever more complex problems, many of which can be formulated as generate-and-test optimization problems. General-purpose optimization algorithms are not well suited for real-world scenarios where many instances of the same problem class need to be repeatedly and efficiently solved, such as routing vehicles over highways with constantly changing traffic flows, because they are not targeted to a particular scenario. Hyper-heuristics automate the design of algorithms to create a custom algorithm for a particular scenario.

Hyper-heuristics typically employ Genetic Programming (GP) and this project has investigated the relationship between the choice of GP and performance in Hyper-heuristics. Results are presented demonstrating the existence of problems for which there is a statistically significant performance differential between the use of different types of GP. Also, some preliminary findings are presented on how to match the type of GP employed in a hyper-heuristic with the problem being addressed.

Biography

Travis is currently a senior in both Computer Science and Computer Engineering. He is an Undergraduate Research Assistant in the Natural Computation Laboratory. He will graduate May 2015 and begin working at Deere & Company as a Software Engineer Summer 2015.

Research Category

Sciences

Presentation Type

Poster Presentation

Document Type

Poster

Location

Upper Atrium/Hall

Presentation Date

15 Apr 2015, 9:00 am - 11:45 am

Comments

Joint project with Sean Harris

This document is currently not available here.

Share

COinS
 
Apr 15th, 9:00 AM Apr 15th, 11:45 AM

Matching Hyper-heuristics and Genetic Programming

Upper Atrium/Hall

Modern society is faced with ever more complex problems, many of which can be formulated as generate-and-test optimization problems. General-purpose optimization algorithms are not well suited for real-world scenarios where many instances of the same problem class need to be repeatedly and efficiently solved, such as routing vehicles over highways with constantly changing traffic flows, because they are not targeted to a particular scenario. Hyper-heuristics automate the design of algorithms to create a custom algorithm for a particular scenario.

Hyper-heuristics typically employ Genetic Programming (GP) and this project has investigated the relationship between the choice of GP and performance in Hyper-heuristics. Results are presented demonstrating the existence of problems for which there is a statistically significant performance differential between the use of different types of GP. Also, some preliminary findings are presented on how to match the type of GP employed in a hyper-heuristic with the problem being addressed.