Friday, November 28, 2008

Pattarawit Polpinit PhD defense

Congratulations to my PhD student Pattarawit Polpinit, who today successfully completed his PhD defense, and passes subject to the usual minor modifications. The title of the thesis is "The Computation of Equilibria in Congestion Networks". Piotr Krysta was internal examiner, Tom Friedetzky was external.

Thursday, November 06, 2008

A fun partisan statistic

This article at the BBC news web site points out that stock markets do better under Democratic administrations than Republican ones. While it seems like the evidence for this is genuine, the following observation (taken from the article) probably pushes it too far:

"Indeed, a recent study published in the New York Times showed that $10,000 (£6,263) invested in the S&P 500 in 1929 would have grown to $11,733 if invested under Republican presidents, but to $300,671 under Democratic presidents."

It sounds like a fun fact to point out to a US voter with a defined-contribution pension scheme. But, I note that it was Republican president Herbert Hoover who presided over the 1929-33 Depression, so perhaps the Democrats had a bit of a headstart in this race. (Meanwhile, share prices right now are still doing badly. But I guess Obama hasn't been sworn in just yet.)

Wednesday, November 05, 2008

US election result

It is known (see this article) that foreigners (to the USA) strongly preferred Obama to McCain. Also, plenty of anecdotal evidence suggests that academics generally strongly prefer Obama. Being a non-USA academic, it should come as no surprise that I too was backing the winner.

A purist might note that the above conclusion makes some sort of statistical assumption along the lines of "naive Bayes" that two distinct indicators for an outcome should independently add to our belief in that outcome. And as it happens, that's a good observation in this case -- for foreign academics there is a potential downside to the Obama victory.

The problem is easy to state. The academic marketplace, both for students and faculty, has become increasingly globalised. And in the world's leading English-speaking nation, the conditions look much improved for it to become a magnet for students and staff alike. When it comes to recruiting overseas students, and retaining staff, the UK and Australia have maybe had a relatively easy ride over the last few years.

Still, I supported Obama anyway, and here is the "professional" reason. Few academics can afford to turn their backs on the USA. And the USA, with its dominance of the algorithms and computational complexity scene, is extra important to those of us in that field. The European algorithms community is better-off for having their American counterparts in a nation that is willing to work with and cooperate with the rest of the world.

Monday, November 03, 2008

Peer review survey

Here is a link to an online survey being conducted by EPSRC, soliciting feedback on their peer review process, used for research grant proposals. Dissatisfied as I am with this process, I felt the urge to contribute. (The survey's closing date is November 7th.)

The general problem with their procedure for peer review, is that it is very primitive, in comparison with the process by which conference submissions are reviewed. You send in a grant application, you get 3 or 4 reviews, you get to submit a response, and these get considered by the panel. A specific problem with this is the shortage of interaction amongst the parties. Suppose that two reviewers think the proposal is excellent, and another thinks it's rubbish. The correct thing to do, surely, is to get the reviewers to settle their differences via a discussion, and preferably solicit another review. This is exactly what program committees do when papers are selected for conferences. In some conferences that I have been involved with, the authors of a submitted paper make a response to the reviews, the reviewers then continue their discussion in the light of the author's response. The whole thing is much more interactive, and helps to ensure transparency and accountability.

Given that it is more significant to get a research grant than to publish a paper (both for the grant applicant/author, and for the taxpayer) it is bizarre in the extreme that EPSRC still uses such a primitive system. There is a case for still more effort being directed towards ensuring that good choices are made. For example, perhaps research grant applicants should be interviewed by the panel (however, suggestions like that go beyond the scope of the above mentioned survey).