CABOB: a fast optimal algorithm for combinatorial auctions

  • Authors:
  • Tuomas Sandholm;Subhash Suri;Andrew Gilpin;David Levine

  • Affiliations:
  • Computer Science Department, Carnegie Mellon University, Pittsburgh, PA;Department of Computer Science, University of California, Santa Barbara, CA;CombineNet, Inc., Pittsburgh, PA;CombineNet, Inc., Pittsburgh, PA

  • Venue:
  • IJCAI'01 Proceedings of the 17th international joint conference on Artificial intelligence - Volume 2
  • Year:
  • 2001

Quantified Score

Hi-index 0.00

Visualization

Abstract

Combinatorial auctions where bidders can bid on bundles of items can lead to more economical allocations, but determining the winners is NP-complete and inapproximable. We present CABOB, a sophisticated search algorithm for the problem. It uses decomposition techniques, upper and lower bounding (also across components), elaborate and dynamically chosen bid ordering heuristics, and a host of structural observations. Experiments against CPLEX 7.0 show that CABOB is usually faster, never drastically slower, and in many cases drastically faster. We also uncover interesting aspects of the problem itself. First, the problems with short bids that were hard for the first-generation of specialized algorithms are easy. Second, almost all of the CATS distributions are easy, and become easier with more bids. Third, we test a number of random restart strategies, and show that they do not help on this problem because the run-time distribution does not have a heavy tail (at least not for CABOB).