Kareem Amin, Michael Kearns, Peter Key, and Anton Schwaighofer
We consider the budget optimization problem faced by an advertiser participating
in repeated sponsored search auctions, seeking to maximize the number of clicks attained under that budget.
We cast the budget optimization problem as a Markov Decision Process (MDP) with censored observations, and propose a learning algorithm based on the well-known Kaplan-Meier or product-limit estimator. We validate the performance of this algorithm by comparing it to several others on a large set of search auction data from Microsoft adCenter, demonstrating fast convergence to optimal performance.
In Uncertainty in Artificial Intelligence (UAI)
Publisher Uncertainty in Artificial Intelligence (UAI)