We consider the problem of designing a profit-maximizing single-item auction, where the valuations of bidders are correlated. We revisit the k-lookahead auction introduced by Ronen [6] and recently further developed by Dobzinski, Fu and Kleinberg [2]. By a more delicate analysis, we show that the k-lookahead auction can guarantee at least e1-1/k/e 1-1/k+1of the optimal revenue, improving the previous best results of 2k-1/3k-1 in [2]. The 2-lookahead auction is of particular interest since it can be derandomized [2, 5]. Therefore, our result implies a polynomial time deterministic truthful mechanism with a ratio of√e/√e+1 ≈ 0.622 for any single-item correlated-bids auction, improving the previous best ratio of 0.6. Interestingly, we can show that our analysis for 2-lookahead is tight. As a byproduct, a theoretical implication of our result is that the gap between the revenues of the optimal deterministically truthful and truthful-in- expectation mechanisms is at most a factor of 1+√e/√e. This improves the previous best factor 5/3 of in [2]. © 2011 Springer-Verlag Berlin Heidelberg.
CITATION STYLE
Chen, X., Hu, G., Lu, P., & Wang, L. (2011). On the approximation ratio of k-lookahead auction. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 7090 LNCS, pp. 61–71). Springer Verlag. https://doi.org/10.1007/978-3-642-25510-6_6
Mendeley helps you to discover research relevant for your work.