A Test Suite for Combinatorial Auctions

Many researchers have proposed algorithms for determining the winners of general combinatorial auctions, with encouraging results. (Some of these algorithms are described in Chapter 14 of this book.) This line of research has given rise to another problem, however. In order to evaluate—and thus to improve—such algorithms, it is necessary to use some sort of test data. Unfortunately, there is little data recording the behavior of real bidders upon which such test data may be built; furthermore, even as such data becomes available, algorithmic testing will require benchmarks which permit arbitrary scaling in the problem size. It is thus necessary to generate generate artificial data that is representative of the sort of scenarios one is likely to encounter.