Research in Online Ad Markets: Automation, Robustness, and Learning
Abstract
We provide an overview of recent trends and future challenges in online ad markets with a focus on robustness, incentive-aware learning, and the auto-bidding world. In the first part, we survey recent results and highlight challenges that emerge as the majority of advertisers adopt automatic bidding algorithms to bid on their behalf, and the market transitions to the so-called auto-bidding world. As part of the move to more automation, we discuss (adversarial) robustness in designing online algorithms, learning, and pricing schemes. In particular, we present online learning algorithms that work under stochastic models with adversarial corruptions and describe recently developed algorithmic techniques that achieve the best of many worlds in a variety of online stochastic and adversarial models. Along the same lines, we discuss the design of dynamic mechanisms for repeated auctions that are robust to the presence of noise or imprecise forecasts. Next, we touch upon robustness in pricing and explore incentive-aware learning algorithms for pricing in both dynamic and static auctions. Finally, we complement these results by exploring data-driven techniques to measure the extent to which bidders have incentives to strategize their bids in ad auctions.
Bio
Vahab Mirrokni is a distinguished scientist, serving as senior director for the New York and Zurich algorithms research groups at Google Research. He received his Ph.D. from MIT in 2005 and his B.Sc. from Sharif University of Technology in 2001. He joined Google Research in 2008, after spending a couple of years at Microsoft Research, MIT, and Amazon.com. His research areas include algorithms, distributed and stochastic optimization, and computational economics, focusing on market algorithms, graph mining, and large-scale optimization.
