I just got back from the International Symposium on Information Theory, held in Seattle. This is the premier conference on Information Theory and is sponsored by the Information Theory Society in the IEEE. Overall I was very impressed with the conference. It is bigger than a workshop but smaller than GLOBECOM or ICASSP. There were around 8 parallel sessions at once.

I was especially impressed with the presentations at the conference. While the Transactions on Information Theory these days seems to glorify obtuse writing and an overly mathematical exposition, many of the presentations I attended did not reflect this. While some presented the utmost details of their proofs to a sleeping audience, quite a few had a nice balance between theory and intuition. I was especially impressed by the asymptotic results. For example, Mike Honig had a nice result on limited feedback systems with training, including relationships between capacity, codebook size, the length of training signal, and the fraction of data.

Overall I think it would be worthwhile to spend more time thinking about system level issues from an information theoretic framework.