Human-Guided Tabu Search

We present a human-guidable and general tabu search algorithm. Our work expands on previous interactive optimization techniques that provide for substantial human control over a simple, exhaustive search algorithm. User experiments in four domains confirm that human guidance can improve the performance of tabu search and that people obtain superior results by guiding a tabu algorithm than by guiding an exhaustive algorithm.

Authors

• Gunnar W. Klau Vienna University of Technology; Neal Lesh and Joe Marks
• Mitsubishi Electric Research Laboratories; Michael Mitzenmacher Harvard University

Keywords

No keywords are indexed for this paper.