Graduate students regulate their circadian clock, blood sugar level, and mating periods according to conferences deadlines. For robotics people, the revered date is not September 21 (solstice), but September 15, the deadline for ICRA. For control people, the date is September 15… or 22… or 29, the ever-slipping deadline for ACC.

The peaks of my productivity for the summer were:

• Getting a driving license. It took 2 tries and too much good money spent on driving lessons.
• Understanding what’s the use of all the buttons on my camera, including the elusive “AEL”.
• Finishing reading the last two books of the Hyperion Cantos. (I’m actually disappointed because Simmons doesn’t explain the origin of the Shrike.)

My good propositions for the next academic year are the same as ever:

• I will buy a car,
• I will read all the papers in my giant “INBOX” folder,
• I will go swimming every other day,
• I will take at least two classes,
• I will write more often on this blog,

and on, and on, and on: I consider myself lucky to be in harmony with my procrastination.

I start with the easy one, as opposed to the difficult one. First thing in the morning I read the paper again. I don’t like it. If I were to write it again, I’d write it in a completely different way. I read the reviews, and it seems that the reviewers liked it more than I like it now. The interesting thing is that I realize this is a case of cognitive dissonance, because when I read the reviews for the first time a couple of months ago, I had the impression they were particularly harsh. Instead they are quite reasonable and helpful.

I blame it on the typical emotional roller-coaster for research, which is conveniently summarized below:

The emotional roller-coaster for engineering papers

• T₀, idea: this is so cool!
• T₀ + 1 month, implementation: wait, it’s not that cool!
• T₀ + 2 months, writing: yeah, it’s cool enough!
• T₀ + 3 months, deadline: enough with this stuff!
• T₀ + 3 months + 1, the day after the deadline: my best paper ever!
• T₀ + 6 months, reading the reviews: nobody understands me!
• T₀ + 8 months, preparing the final version: this paper sucks, let me rewrite it from scratch!
• T₀ + 10 months, conference presentation: how can I fake excitement for an idea I had 1 year ago?
• T₀ + 1 year, considering writing a journal version: I don’t have time for this old stuff, let’s do something new!
• T₀ + 2 years, casually opening the PDF: that’s really cool! did I write that?!

This is specific to engineering, as science and math have very different psychology. Hopefully I will be doing a bit of biology in the near future, and so I’ll be able to report the differences. For sure, it will not take just 3 months from the idea to the writing.

“Historical contingency and the evolution of a key innovation in an experimental population of Escherichia coli”

doi:10.1073/pnas.0803151105

The Long Term Evolution Experiment has been growing 12 colonies of E. Coli in controlled conditions since 1988. At one point, after 30000 generations, one of the colonies evolved the ability to synthetize citrate in addition to glucose. Because the researchers had frozen samples from previous generations, they could restart the evolution process from an earlier point and see it happen again.

And no, I’m not doing biology here at Caltech. I would never wait 20 years for my experiments to finish!

In his opinion, the currently hot topics are energy and climate change. To which I replied: what about curing cancer? He told me that if one finds a cure for some kind of cancer, at best he might help, say, a million people, but if one solves either of those two problems, he will help all of humanity, and perhaps save some hundred million people that one day will die due to wars and chaos caused by the shortage of energy and the climate change.

He then told me that a good reference for choosing a topic is to visualize a simplified two-axis diagram. One axis is the novelty of a topic, and the other is its importance. Ideally, one would want to work on a topic which is highly innovative applied to a very important problem. However, realistically, it is easier to make either some incremental contribution to a well-studied important topic (a new technique for improving by 0.5% the efficiency of battery chargers) or to come up with some crazy theory which doesn’t have applications yet.

My impression is that, here in the US, professors have lots of PhD students available, therefore they can diversify the risk and work on lots of high-risk/high-reward projects (just like venture capitalists).

He told me this is a good time to see what’s going on in various disciplines and to talk to people here at Caltech whose interests overlap (even for an epsilon) with mine.

Here’s the list, in no particular order:

• Emmanuel Candes (compressed sampling, detection vs. estimation)
• Christof Koch
• Michael Dickinson
• Houman Owhadi
• Ralph Adolphs (MRI)
• Steven Quartz (decision making)
• Mani Chandy (distributed decision making, super idempotent functions)
• Joel Burdick
• Pietro Perona

Plus, we talked about some of topics being currently studied in his group:

• networks (Shi Ling)
• verification (Nok)
• biological computing (Elisa Franco)

Research, at last! I was starting to accept that my main activity here at Caltech would have been doing homework.