I had the pleasure of attending the NSF Center for Science of Information Site Visit earlier this week. It was a great experience. I met some of the students, post-docs, and professors in the information theory field and I got to learn about some of the projects people were doing.
I particularly liked the discussions on projects by Tom Courtade, a post-doc at Stanford. He introduced a problem on boolean functions and a problem on lossy compression of graphs using the log-loss distortion function. The first project involves finding I(X^n, b(Y^n)) where b is a boolean function of n bits, and Y^n is a noisier version of X^n. The question seems simple, but actually, as Tom explained, it is very complicated. The second problem presented an interesting way of encoding graphs. There were however some disagreements on whether log-loss is a good measurement of distortion for applications of lossy encoding of graphs.
I had the chance to meet many of professors who are at the graduate schools I am applying to and I shared my poster with many of them (click here to see my poster). I was amused when Professor David Tse came by to see my poster. When I told him that I was an undergrad applying to Berkeley, he said to me in a very serious voice, "Consider this your interview". When I finished telling him about my project, he said again in a very serious voice, "I will be reading your application". I guess I should consider that a good thing, since he could have told me that he wasn't going to read my application.
Overall, I think the conference was a great opportunity for me to see what goes on the field of information theory. I now have a better understanding than before on what the research projects are like. I enjoyed the academic atmosphere and I enjoyed meeting the grad students who told me about what grad student life was like.
Unrelated to the conference, today I got a chance to speak with Professor Ruby Lee, who has the background knowledge on cache. When I explained what I was trying to do for my project, she suggested that I run some tests to collect data before making assumptions about the timing distributions of the cache hits and the cache misses. Though I find gathering data to be much less interesting than theory, I guess she makes a good point. I will see what I can do about this. Meanwhile, I will continue to try to see what the Chernoff information can tell us about the error in making a guess about the distribution.
Here is a question from the Putnam exam from last weekend. It dutifully honors the year 2012:
Given any set of 12 numbers selected from the range (1, 12), is it possible to choose three distinct elements from this set which are the lengths of an acute triangle?
I will have a busy next few weeks ahead of me due to grad school apps. Hopefully I will have time to update this blog.
