Incentive-Compatible Learning of Reserve Prices for Repeated Auctions
研究了拍卖方如何从历史出价中学习保留价以优化收入，并考虑投标人的策略行为；发现使用个性化保留价可接近最优收入。
How can an auctioneer optimize revenue by learning the reserve prices from the bids in the previous auctions? How should the long-term incentives and strategic behavior of the bidders be taken into account? Motivated in part by applications in online advertising, in “Incentive-Compatible Learning of Reserve Prices for Repeated Auctions,” Kanoria and Nazerzadeh investigate these questions. They show that if a seller attempts to dynamically update a common reserve price using the bidding history, buyers will shade their bids, which can hurt the revenue. However, when there is more than one buyer, using personalized reserve prices, the auctioneer can achieve a near-optimal revenue. In their proposed mechanism, the personal reserve price for each buyer is determined using the historical bids of other buyers.