Learning Reserve Prices in Second-Price Auctions
summary
Proves that learning an optimal anonymous reserve for second-price auctions needs nearly-tight sample complexity equal to prior Myerson bounds while crucially not depending on the number of bidders. Builds an algorithm that extracts first and second highest samples, constructs dominated empirical CDFs and an analytic doubly-shaded CDF, then returns the reserve maximizing revenue on the dominated model. Analysis uses revenue-monotonicity, extreme-value theorems and tail bounds for regular and MHR classes, and holds for correlated values under bounded-support.
AI-written summary · about
chapters (5)
- 0:00Introduction to auctions
- 11:02Literature and results
- 14:15Algorithm description
- 18:46Analysis and proofs
- 28:18Technical distribution tools
related talks
48mOutposts Between Worst- and Average Case Analysis: A Case Study in Auction Design
43mLearning about Auctions
15mFundamental Limits Of Learning In Data-Driven Problems
09mMechanism Design Via Machine Learning: Overfitting, Incentives, and Privacy
01h 07mMulti-Dimensional and Non-Linear Mechanism Design and Approximation I
19mMarket Algorithms for Autobidding
27mEconometrics for Learning Agents
35mThe Sample Complexity of Revenue Maximization