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:Computer Laboratory Systems Research Group Seminar
SUMMARY:A Bit of Network Information Theory - Suhas Diggav
 i (EPFL)
DTSTART;TZID=Europe/London:20100112T150000
DTEND;TZID=Europe/London:20100112T160000
UID:TALK22063AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/22063
DESCRIPTION:Network information theory deals with fundamental 
 characterizations of information flow and/or compr
 ession over communication networks. While there ar
 e some satisfying characterizations for flows on n
 etworks represented as graphs\, there have been fe
 w complete characterizations for networks with mor
 e complicated interactions such as those encounter
 ed in wireless networks. Given this discouraging s
 tate of affairs\, a natural question to ask is whe
 ther there are methodologies to make progress on t
 hese questions.\n\nIn this talk we posit that one 
 could gain insight into these problems by asking f
 or less\, i.e.\, by looking for (provable) approxi
 mate characterizations\, with the hope that in ter
 ms of engineering solutions\, these might actually
  be enough. Underlying these approximations are ex
 act results for deterministic/lossless network com
 munications.\n\nWe illustrate this approach by exa
 mining two problems. First is that of information 
 flow over wireless Gaussian relay networks\, where
  we show that the achievable rate is within a cons
 tant number of bits from the information-theoretic
  cut-set upper bound on the capacity of these netw
 orks. Underlying this result is an exact informati
 on-theoretic max-flow min-cut characterization of 
 a deterministic relay network. The second is the a
 pproximate characterization of the K-description G
 aussian multiple description source coding problem
 .\nHere we show that the rate region can be sandwi
 ched between two polytopes with matching hyperplan
 es\, separated by at most one bit. Underlying this
  is an exact characterization of a lossless multi-
 level source coding problem.\n\nThese results show
  that characterizations\, within a constant number
  of bits approximation\, may be a promising direct
 ion to get insight in network information theory p
 roblems. We will conclude with implications of thi
 s methodology by illustrating several applications
  such as wireless network secrecy.\n\nParts of thi
 s work were done in collaboration with S. Avestime
 hr (Cornell)\, S. Mohajer (EPFL)\, C. Tian (AT&T) 
 and D. Tse (UC\, Berkeley).\n\nBio: Suhas N. Digga
 vi received the B. Tech. degree in electrical engi
 neering from the Indian Institute of Technology\, 
 Delhi\, India\, and the Ph.D. degree in electrical
  engineering from Stanford University\, Stanford\,
  CA\, in December 1998.\n\nAfter completing his Ph
 .D.\, he joined the Information Sciences Center\, 
 AT&T Shannon Laboratories\, Florham Park\, NJ\, wh
 ere he was a Principal Member Technical Staff. He 
 is currently on the faculty of the School of Compu
 ter and Communication Sciences\, EPFL\, where he h
 eads the Laboratory for Information and Communicat
 ion Systems (LICOS). He will join the University o
 f California\, Los Angeles as a full professor in 
 2010. His research interests include information t
 heory\, network communications\, data compression 
 and signal processing. More information can be fou
 nd at http://licos.epfl.ch.\n\nHe is a recipient o
 f the 2006 IEEE Donald Fink prize paper award\, 20
 05 IEEE Vehicular Technology Conference best paper
  award and the Okawa foundation research award. He
  is currently an Associate Editor for the IEEE/ACM
  Transactions on Networking and the IEEE Transacti
 ons on Information Theory. He has served as an ass
 ociate editor for the IEEE Communication Letters\,
  a guest editor for IEEE Selected Topics in Signal
  Processing and on the technical program committee
  for several conferences including ISIT\, ICC\, IT
 W\, and Globecom. He has 8 issued patents.\n
LOCATION:FW26\, Computer Laboratory\, William Gates Builidi
 ng
CONTACT:Eiko Yoneki
END:VEVENT
END:VCALENDAR