Details for Talk on: 09.11.2016

Combinatorial auctions are attractive mechanisms to efficiently allocate resources even when bidders have combinatorial valuations. However, the large number of items sold in many real-world auctions prevents a direct application of combinatorial auction formats. In particular, designing concise bidding languages that do not constrain the bidders' expressiveness is a challenging task. In this talk, we will present a new paradigm for designing bidding languages based on achine learning principles. We show that we can exploit simple knowledge about bidders' preferences to automatically design concise bidding languages that don't limit the expressiveness of the bidders.