But what to write about? It's hard to choose just one topic to make a "post of the year". So, I'll just write about some work-related highlights of 2013, big and small, in no particular order:
- The newly founded Simons Institute for the Theory of Computing seems to have had a great year. I haven't had a chance to visit, but I have watched some videos of their excellent talks online! And I've been keeping up with Moritz Hardt's interesting new blog.
- I've for the first time gotten involved in running a conference, or rather part of it: ISAIM 2014 will feature a session on the "Theory of Machine Learning," and I'm quite happy with the program. (It's not too late to attend!)
- My friend Nikhil Srivastava (who appeared on this blog before), together with Adam Marcus and Dan Spielman, proved the Kadison-Singer conjecture. It's a result that touches on many areas of mathematics -- its resolution a big deal, and especially exciting it came from the computer science community.
- With the opening of SCDA, Facebook Research Labs, MSR-NYC, and others, machine learning seems to be quite in demand, both in academia and industry. (And, not quite unrelatedly, deep learning is continuing to make a comeback.) It's a great time to be a machine learning researcher or practitioner.
- Near the beginning of the year, my co-authors and I finished and published a paper extending statistical queries to optimization algorithms and proved a lower bound for planted clique. I'm particularly happy about this result, which had been in the works for a while.
- I published my first paper with students at UIC. One of them, Jeremy Kun, who is working with me, has a great blog, which you should all read! This also happens to be my first work primarily in game theory, which seems to be an area many learning researchers delve into.
- While last year was 100 years since Alan Turing's birth, this year was Erdős's centenary. One could say that Erdős was the Turing of combinatorics, and he had quite a bit of impact in computer science as well.
- It looks like the NSA hasn't made any huge mathematical breakthroughs, but that hasn't kept them from breaking crypto-systems in the real world, both in ways we'd expect and in ways we wouldn't.
- Leslie Valiant published Probably Approximately Correct. I'm sold, but can it bring learning theory to the masses?