Landscape Analysis and Algorithm Development for Plateau Plagued Search Spaces
Abstract
Over the last 10 to 20 years, heuristic search in the Operations Research and Artificial Intelligence communities has focused on developing high level general purpose algorithms, such as Tabu Search and Genetic Algorithms. However, understanding of when and why these algorithms perform well still lags. Our project extended the theory of certain combinatorial optimization problems to develop analytical characterizations of portions of search spaces and as the basis for creating new algorithms for two well known problems. We focused attention on two specific subclasses of NP-Hard problems: elementary landscapes, which include Traveling Salesman Problem (TSP) and embedded landscapes, which include Maximum Satisfiability (MAXSAT). Our analysis supports calculating exactly the statistical moments of the distributions of values in regions of MAXSAT search spaces and explains why true plateaus are rare. Our new algorithm for TSP is competitive with the state of the art, while being much simpler and easier to understand.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 28, 2011
- Accession Number
- ADA547002
Entities
People
- Adele Howe
- L. D. Whitley
Organizations
- Colorado State University