|
|
MIT Electrical Engineering and Computer Science
EECS Event |
Monday, April 23, 2001
4:00 PM (refreshments 3:45)
Grier Room, Room 34-401A
EECS Special Seminar
Abstract
Coding Problems in Communication Networks.
Abstract:
The field of network information theory deals with capacity, coding, and control problems that arise in the context of communication systems involving multiple senders and/or receivers. Unlike for the information theory of single-user channels however (reasonably well understood by now, at least in many important special cases), a complete theory of information in networks still lies beyond our grasp, and the field remains largely unexplored. A few results are known for some special networks (such as broadcast channels, multiple access channels, distributed sources), but even in these cases, much less is known about practical code constructions than about information-theoretic performance bounds. Yet, despite this apparent lack of a solid foundation, and fueled primarily by the success of the Internet in recent years --and also by its expected impact on our future economy--, network channels and distributed information systems are becoming two of the most important topics of research in modern communications. In this talk I will present our work on network coding problems. We will start with a quick look at the evolution of packet networks (starting with the old ARPANET), with the goal of setting up a context in which to discuss where and how network codes can and should play an important role. Then I will present an overview on a number of activities we have currently under way in this regard, with an emphasis on two specific cases: (a) reliable communication over a covert packet channel, and (b) applications of multiple description codes to various problems of video transmission over IP networks. To conclude, we will briefly discuss a few examples of what I believe are very interesting and challenging open problems in this area.