Sciweavers

AAAI
2010

Stability and Incentive Compatibility in a Kernel-Based Combinatorial Auction

13 years 11 months ago
Stability and Incentive Compatibility in a Kernel-Based Combinatorial Auction
We present the design and analysis of an approximately incentive-compatible combinatorial auction. In just a single run, the auction is able to extract enough value information from bidders to compute approximate truth-inducing payments. This stands in contrast to current auction designs that need to repeat the allocation computation as many times as there are bidders to achieve incentive compatibility. The auction is formulated as a kernel method, which allows for flexibility in choosing the price structure via a kernel function. Our main result characterizes the extent to which our auction is incentive-compatible in terms of the complexity of the chosen kernel function. Our analysis of the auction's properties is based on novel insights connecting the notion of stability in statistical learning theory to that of universal competitive equilibrium in the auction literature.
Sébastien Lahaie
Added 29 Oct 2010
Updated 29 Oct 2010
Type Conference
Year 2010
Where AAAI
Authors Sébastien Lahaie
Comments (0)