Monday, June 29, 2009

ISIT conference this week

The 2009 International Symposium on Information Theory is going on this week in Seoul, Korea. I'm not attending, although it sounds like there's plenty of interesting things going on. Amin Shokrollahi is giving a tutorial on fountain codes, Balaji Prabhakar is talking about models and algorithms for Internet data centers. Some of the plenary talks sound like they'd be of interest to CS folks: Randomized Dimensionality Reduction by Richard Baraniuk, It's Easier to Approximate by David Tse, Combinatorial Reasoning in Information Theory by Noga Alon. Plenty of sessions on low-density parity-check codes and their variations, and network coding. There's something relatively new out there called "Polar codes" that I should learn more about.

It's a bit disappointing to see in the program a lack of what I think of as "CS theorists" at ISIT, though perhaps this year Seoul was just too far to go for a conference that's more tangential to people working on related areas in this group. As I've said before, given the list of topics, one would think there would be more crossover. With eight parallel sessions over 4 1/2 days, there would certainly seem to be room. Maybe next year.

If anyone in the comments wants to point out exciting news or results from ISIT, or other blogs discussing the goings-on, please do so; I'd be happy to hear of news.


Anonymous said...

Theoretical CS already has its two top tier conferences (FOCS and STOC) which accepts between themselves 150+ papers. On top of them is the plan to have an even more elite conference (namely ICS).
Then there are several secondary (but still mainstream) conferences such as SODA/SoCG/ICALP/Complexity etc.
I do not think there is really enough people working in the core areas of theoretical CS to produce enough publishable papers per year to fill all the (thick) proceedings of these conferences. (I am not counting papers based on applications of ideas developed in TCS, such as in economics/game theory/networks etc.)

So no wonder people are reluctant to submit to non-mainstream conferences such as ITIS.

Michael Mitzenmacher said...

Anonymous #1: I suspect your comment is based on ignorance, so let me clarify.

ISIT is the BIG annual conference on information theory. It is not only mainstream for people in that area, it's a conference where you can expect to see most people in the field.

I'd agree that theory CS people, naturally, tend to submit to the many theory CS conferences. But if you're doing a topic connected to information theory -- coding, compression, network coding, compressed sensing, estimation, etc. you'll get a much more closely connected audience at ISIT than at FOCS/STOC. Many TCS people work in these areas, and occasionally you see them with work in ISIT. Such a crossover clearly benefits both subfields.

asarwate said...

I generally blog about ISIT each year over at my blog. Hopefully this year I'll be a bit more timely about things.

Andrew Eckford said...

Reporting live from the COEX in Seoul ...

Good conference so far. Most of the ISITs that I've attended have had an unofficial "theme", reflecting the contemporary research emphasis in the Info Theory community. This time there doesn't seem to be one as such -- it's been very well balanced.

A lot of people were excited about Baraniuk's excellent plenary, so compressed sensing looks like the next big thing.

Anonymous said...

Find Allerton more travel friendly, usually meet most of the ISIT crowd there and more.

Unknown said...

My area of expertise is quite narrow, but for me at least, it seemed a somewhat less interesting ISIT than in other years. More people seemed to skip it this year than usual, judging by anecdotal evidence. I also get the feeling mathematical results are more well received than algorithmic results, which doesn't help with the lack of algorithm folks. Many people believe this needs to change, but that's easier said than done. When I asked my associates on Friday whether there were any revelatory results, the strongest answer I got was that the Shannon Lecture had some; most years that's the last place you'd expect to find them.

Still, it's always good to see all the familiar faces and to know that someone's going to see your work; you just have to have a presentation that makes sure they understand it as well.