| | |
Summary: Ordered Greed
Peter G. Anderson and William T. Gustafson
Rochester Institute of Technology
Rochester, NY 14623-0887
anderson(at sign)cs.rit.edu
April 10, 2000
Abstract
Scheduling problems are among the most challenging and realis-
tic problems application of problem solving heuristics, such as genetic
algorithms (GAs). The naive greedy algorithm for scheduling simply
assigns, in turn, each item to be scheduled the best yet untaken posi-
tion for that item. We investigate using a genetic algorithm to search
the space of orderings for this greedy algorithm. That is, the GA
individuals are permutations that determine the permutations that
are the schedules, rather than the GA individuals directly being the
schedules.
We have experimented with the classical N Queens problem and
a realistic soccer tournament scheduling problem, comparing the GA
individual as the assignment with our greedy hybrid algorithm (\or-
dered greed").
|
