Computational Mechanism Design Optimal Auction Design

In a single experiment all agents have the same and c

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: roblem Fig 3 compares the e ciency a, and revenue b, for markets with N = 10 agents as the e ectiveness 1 ,  and cost C of deliberation changes. In a single experiment all agents have the same and C parameters. Equivalently, we model agents with the same bounded resources, but easy local problems for large 1 ,  and relaxed time-constraints for small costs C . 9 C = 0.05 100 90 100 90 C = 0.1 100 90 C = 0.05 100 90 C = 0.1 Efficiency (%) Efficiency (%) Revenue (%) 80 70 60 50 40 0 0.2 0.4 0.6 0.8 Computational Effectiveness, 1 − α C = 0.5 100 90 80 70 60 50 40 0 80 70 60 50 Revenue (%) 0 0.2 0.4 0.6 0.8 Computational Effectiveness, 1 − α C = 0.5 80 70 60 50 40 0 0.2 0.4 0.6 0.8 Computational Effectiveness, 1 − α C=1 100 90 0.2 0.4 0.6 0.8 Computational Effectiveness, 1 − α C=1 40 100 90 100 90 Efficiency (%) Efficiency (%) Revenue (%) 80 70 60 50 40 0 0.2 0.4 0.6 0.8 Computational Effectiveness, 1 − α 80 70 60 50 40 0 80 70 60 50 Revenue (%) 0 0.2 0.4 0.6 0.8 Computational Effectiveness, 1 − α 80 70 60 50 40 0 0.2 0.4 0.6 0.8 Computational Effectiveness, 1 − α a b Fig. 3. Auction performance for N = 10 agents with di erent levels of deliberation e ectiveness and cost. `+' ascending-price; `x' sealed-bid; `o' posted-price sequential auctions. Bid increment = 0:4, 1000 trials. a E ciency; b Revenue. 0.2 0.4 0.6 0.8 Computational Effectiveness, 1 − α 40 In this market, with N = 10 agents, when agents have small deliberation costs C = 0:05 then A S P , and A  S P for C = 0:1. Auctions with dynamic prices, such as the ascending-price or sealed-bid auctions, perform better than the posted-price auction for small C . For medium deliberation costs C = 0:5, the sealed-bid auction fails and A  P S . For large deliberation costs e.g. C = 1 there is no deliberation in any auction, and auction design does not matter. E ciency and revenue increase as deliberation e ectiveness increases, and decrease as deliberation cost increases. The ascending-price and sealed...
View Full Document

This document was uploaded on 03/19/2014 for the course COMP CS286r at Harvard.

Ask a homework question - tutors are online