| The William E. (Gene) Brice Colloquium Series: Feedback, Erasures, Zero-Errors, and More James L Massey, Professor of Digital Systems Engineering, The Swiss Federal Institute of Technology (ETH), Zurich, Switzerland | |
| date: | 4:00PM to 5:00PM US Central (GMT −0600) Thursday, January 26, 2006 |
| length: | 1 hour, 0 minutes |
| location: | McMurtry Auditorium Duncan Hall |
| sponsor: | Electrical and Computer Engineering |
| summary: | Claude E. Shannon devoted the inaugral "Shannon Lecture" at the 1973 IEEE International Symposium on Information Theory (ISIT) to the role of feedback in communications, implicitly challenging other information theorists to get to work on this topic. Peter Elias based his Shannon Lecture at the 1977 ISIT on the binary erasure channel (in which a transmitted binary digit is either received error-free or as an erasure symbol), which he described as the simplest channel that incorporates the essential features of "noisiness". Our aim in this lecture is to take up Shannon's challenge by following Elias's suggestion that the binary erasure channel is a good place to start. We will be led naturally to consider data transmission systems that yield zero-errors, i.e., not just a low probability of error but no errors at all. Shannon already in 1956 defined the zero-error capacity of a channel as the upper limit of rates at which zero-error transmission is possible, but paradoxes that occur when this definition is applied to the binary erasure channel will force us to expand this concept. Elias in 1955 introduced infinite-constraint- length codes in a paper that gave birth to "convolutional codes". We find that new paradoxes occur for zero-error transmission when infinite-constraint-length codes are allowed. Resolving all these paradoxes leads to a taxonomy of channels in which the binary erasure channel occupies a prominent place.
|
| more info: | Event Website Brice Colloquium Series ECE Department |
| watch: | To watch or download the video of the event, click on one of the icons below: |