# ITA Workshop Aftermath

The 2010 ITA Workshop is over, and now that I have almost caught up on sleep, I can give a short report. I think this year was the largest workshop so far, with over 300 speakers.

One of the additions this year was a poster session to give those students who didn’t get a chance to speak at the “Graduation Day” talks on Wednesday an opportunity to present their research. The posters went up on Monday and many stayed up most of the week. I am usually dubious of poster sessions for more theoretical topics; from past experience I have had a hard time conveying anything useful in the poster and they are poorly attended. However, this poster session seemed to be a rousing success, and I hope they continue to do it in future years.

The theme this year was “networks,” broadly construed, and although I didn’t end up going to a lot of the “pure” networking sessions, the variety of topics was nice to see, from learning in networks to network models. Jon Kleinberg gave a great plenary lecture on cascading phenomena. I particularly enjoyed Mor Harchol-Balter’s tutorial on job scheduling in server farms. I learned to re-adjust my intuition for what good scheduling policies might be. The tutorials should be online sometime and I’ll post links when that happens.

The “Senseless Decompression” session organized by Rudi Urbanke and Ubli Mitra should also be online sometime. Apparently 20+ schools contributed short videos on the theme of $\frac{1}{2} \log(1 + \mathrm{SNR})$. Maybe we can make a YouTube channel for them.

Perhaps later I’ll touch on specific talks that I went to, but this time around I didn’t take too many notes, probably because I was a little exhausted from the organizing. I may post a more technical thing on the talk I gave about my recent work on privacy-preserving machine learning, but that will have to wait a bit in the queue. Mor’s tutorial suggests I should use Shortest Remaining Processing Time (SRPT) to make sure things don’t wait too long, and I have some low-hanging tasks that I can dispatch first.