BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Microsoft Research Cambridge\, public talks
SUMMARY:Mean Field Equilibria of Dynamic Auctions with Lea
rning - Krishnamurthy Iyer\, Stanford University
DTSTART;TZID=Europe/London:20120403T090000
DTEND;TZID=Europe/London:20120403T100000
UID:TALK37111AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/37111
DESCRIPTION:We study learning in a dynamic setting where ident
ical copies of a good are sold over time through a
sequence of second price auctions. Each agent in
the market has an *unknown* independent private va
luation which determines the distribution of the r
eward she obtains from the good\; for example\, in
sponsored search settings\, advertisers may initi
ally be unsure of the value of a click. Though the
induced dynamic game is complex\, we simplify ana
lysis of the market using an approximation methodo
logy known as *mean field equilibrium* (MFE). The
methodology assumes that agents optimize only with
respect to long run average estimates of the dist
ribution of other players' bids. We show a remark
able fact: in a mean field equilibrium\, the agent
has an optimal strategy where she bids truthfully
according to a *conjoint valuation*. \nThe conjoi
nt valuation is the sum of her current expected va
luation\, together with an overbid amount that is
exactly the expected marginal benefit to one addit
ional observation about her true private valuation
. \nUnder mild conditions on the model\, we show t
hat an MFE exists\, and that it is a good approxim
ation to a rational agent's behavior as the number
of agents increases. Formally\, if every agent ex
cept one follows the MFE strategy\, then the remai
ning agent's loss on playing the MFE strategy conv
erges to zero as the number of agents in the marke
t increases. We conclude by discussing the implica
tions of the auction format and design on the auct
ioneer's revenue. In particular\, we establish a
dynamic version of the revenue equivalence theorem
\, and discuss optimal selection of reserve prices
in dynamic auctions.\n\n(Joint work with Ramesh J
ohari and Mukund Sundararajan)
LOCATION:Small lecture theatre\, Microsoft Research Ltd\, 7
J J Thomson Avenue (Off Madingley Road)\, Cambrid
ge
CONTACT:Microsoft Research Cambridge Talks Admins
END:VEVENT
END:VCALENDAR