Prof. Marceau Coupechoux
Title: Monte Carlo Tree Search for Auctions
Bio:
Marceau Coupechoux received the Engineer degree from Telecom Paris in 1999 and University of Stuttgart in 2000, the Ph.D. degree from Institut Eurecom in 2004, the Habilitation degree from University Pierre et Marie Curie in 2015. He is a Professor at Telecom Paris and a Professeur Chargé de Cours at Ecole Polytechnique. From 2000 to 2005, he was with Alcatel-Lucent Bell Labs. In 2011-2012, he was a Visiting Scientist with the Indian Institute of Science, Bengaluru, India. He has been a General Co-Chair of WiOpt 2017 and GameNets 2019. In the Computer and Network Science Department of Telecom Paris, he is working on wireless and cellular networks, focusing mainly on performance evaluation, optimization and resource allocation.
Abstract:
Simultaneous Ascending Auctions (SAA) has been used for long time for spectrum allocation. Recent 5G licenses allocations relied last year in several countries on this mechanism. SAA has been developed by Milgrom and Wilson who received the Nobel prize in Economy for their contributions in auction theory in 2020. The success of SAA probably comes from its relative simplicity: although it is multiple items auction, it is not combinatorial. However, this leaves a large space of strategies for the bidders. SAA belongs to the class of multiplayer non-zero sum stochastic large games with partial information for which optimal bidding strategies are unknown and no generic algorithm has yet been designed for solving it. In this talk, we will present a bidding strategy based on Monte Carlo Tree Search that tackles two important strategical challenges that arise with SAA: the exposure problem and the own price effect.