Non-parametric Revenue Optimization for Generalized Second Price Auctions
This addresses revenue optimization for auction platforms, offering incremental improvements with new algorithms and theoretical insights.
The paper tackles the problem of learning optimal reserve prices for generalized second price auctions, introducing a novel algorithm with a running-time complexity of O(n S log(n S)) and providing theoretical guarantees that improve upon prior work, including convergence analysis of empirical equilibrium bidding functions.
We present an extensive analysis of the key problem of learning optimal reserve prices for generalized second price auctions. We describe two algorithms for this task: one based on density estimation, and a novel algorithm benefiting from solid theoretical guarantees and with a very favorable running-time complexity of $O(n S \log (n S))$, where $n$ is the sample size and $S$ the number of slots. Our theoretical guarantees are more favorable than those previously presented in the literature. Additionally, we show that even if bidders do not play at an equilibrium, our second algorithm is still well defined and minimizes a quantity of interest. To our knowledge, this is the first attempt to apply learning algorithms to the problem of reserve price optimization in GSP auctions. Finally, we present the first convergence analysis of empirical equilibrium bidding functions to the unique symmetric Bayesian-Nash equilibrium of a GSP.