The defense/talk went pretty well, I thought. And there was champagne afterwards! However, filing the thesis has been pushed off until early in the summer. Although this has the benefit of making the finished document much better, the downside is that such a large object sitting in my brain tends to crowd out other more exciting projects that I would like to work on or think about. In the meantime, I will try to blog a bit more frequently so that I don’t end up with tunnel vision.
Robin sent along a link to some mathematical models in metal. They look like more hard-core versions of the sculpted surfaces seen in display cases in math department hallways. I took a few math classes at UIUC during high school and I remember walking past these dusty shelves filled with plaster (?) shapes with intriguing names like “the ring of the nodoid.”
This is why I’ve not been posting, but hopefully that will change.
Beyond the ABCs of AVCs : robust and adaptive strategies for future communication systems
Anand D. Sarwate
Advisor : Michael Gastpar
Department of Electrical Engineering and Computer Sciences
University of California, Berkeley
Thursday, May 15
521 Cory Hall
Cutting-edge application areas such as cognitive radio, ad-hoc networks, and sensor networks are changing the way we think about wireless services. The demand for ubiquitous communication and computing requires flexible communication protocols that can operate in a range of conditions. This thesis adopts and extends a mathematical model for these communication systems that accounts for uncertainty and time variation in link qualities. The arbitrarily varying channel (AVC) is an information theoretic channel model that has a time varying state with no statistical description. We assume the state is chosen by an adversarial jammer, reflecting the demand that our constructions work for all state sequences. In this talk I will show how resources such as secret keys, feedback, and side-information can help communication under this kind of uncertainty. I will present results on list coding and rateless coding for discrete channel models and coding with side information for continuous channels.
And of course the most important part: refreshments will be provided!