Information Theory of Bursty Communication
Feb 18, 2014
from 10:00 AM to 12:00 PM
|Where||Engr. IV Bldg., Elliott Room 53-135|
|Contact Name||Prof. Suhas Diggavi|
|Add event to calendar||
The traditional information theoretic model to reliable communication ignores issues related to synchronization and assumes that the receiver is cognizant of the message transmission period. However, in many applications transmission occurs in a bursty fashion with "small" amounts of data transmitted once in a "long" while. For such settings it is unclear whether the cost of acquiring synchronization can be neglected since each burst of data contains a "relatively small” number of bits. What are the limitations due to a lack of synchrony between the transmitter and the receiver in bursty communication? In this talk I will address this question by exhibiting fundamental tradeoffs between the message transmission delay, the energy needed to transmit a message, and the output sampling rate which is taken as a proxy for the receiver’s energy consumption. In particular, we will see that training based schemes, which separately implement message detection and message isolation at the decoder, need not be optimal. I will also discuss a sequential decoder which can sleep almost all the time and yet be as efficient as the best receivers that sample all channel outputs. (Implementing such a decoder for talk attendance will be left as an exercise.)
Aslan Tchamkerten received the B.S. and M.S. degrees in Physics in 2000 and the Ph.D. degree in Communications in 2005, both from EPFL. Between 2005 and 2008 he was a Postdoctoral Associate at MIT. Since 2008 he has been with the Communications and Electronics Department at Telecom ParisTech (ex. Ecole Nationale Supérieure des Télécommunications) where he is currently an Associate Professor. In 2009 he won a junior chair of excellence from the French National Research Agency.