Heavy-Tailed Distributions in Combinatorial Search

Combinatorial search methods often exhibit a large variability in performance. We study the cost pro les of combinatorial search procedures. Our study reveals some intriguing properties of such cost pro les. The distributions are often characterized by very long tails or \heavy tails". We will show that these distributions are best characterized by a general class of distributions that have no moments (i.e., an in nite mean, variance, etc.). Such non-standard distributions have recently been observed in areas as diverse as economics, statistical physics, and geophysics. They are closely related to fractal phenomena, whose study was introduced by Mandelbrot. We believe this is the rst nding of these distributions in a purely computational setting. We also show how random restarts can e ectively eliminate heavy-tailed behavior, thereby dramatically improving the overall performance of a search procedure.