BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Efficient Advert Assignment - Peter Key (MSR Cambridge)
DTSTART:20141125T140000Z
DTEND:20141125T150000Z
UID:TALK55706@talks.cam.ac.uk
CONTACT:Felix Fischer
DESCRIPTION:We develop a framework for the analysis of  large-scale Ad-auc
 tions where adverts are assigned over a continuum of search types. For thi
 s pay-per-click market\, we provide an efficient and highly decomposed mec
 hanism that maximizes social welfare. In particular\, we show that the soc
 ial welfare optimization can be solved in separate optimizations conducted
  on the time-scales relevant to the advertisement platform and advertisers
 . Here\, on each search occurrence\, the platform solves an assignment pro
 blem and\, on a slower time scale\, each advertiser submits a bid which ma
 tches its demand for clicks with supply. Importantly\, knowledge of global
  parameters\, such as the distribution of search terms\, is not required w
 hen separating the problem in this way. This decomposition is implemented 
 in an adversarial setting. Exploiting the information asymmetry between th
 e platform and advertiser\, we describe a simple mechanism which incentivi
 zes truthful bidding and has a unique Nash equilibrium that is socially op
 timal\, and thus implements our decomposition. Further\, we consider model
 s where advertisers adapt their bids smoothly over time\, and prove conver
 gence to the solution that maximizes aggregate utility. Finally\, we descr
 ibe several extensions which illustrate the flexibility and tractability o
 f our framework.\nhttp://arxiv.org/pdf/1404.2750v3.pdf
LOCATION:MR4\, Centre for Mathematical Sciences\, Wilberforce Road\, Cambr
 idge
END:VEVENT
END:VCALENDAR