Woke up, decided not to take a shower, decided to take a shower, took a shower, ate a hasty muesli breakfast, then headed out with Daniel to Algorithmic Game Theory (henceforth referred to as AlgGT).
AlgGT was a three hour lecture, of which the first two hours were spent on basic game theory concepts–I downloaded the textbook and started plowing through it :P But in the third hour the professor, one Paolo Penna, introduced the idea of a potential game. Such a game has an associated potential function that tracks the progress of the game towards equilibrium–everyone’s incremental improvements to their utility (and perhaps detriments to others’ utility) inexorably add into the potential function, which means that no cycles can exist and the game must tend to a single state equilibrium.
At first I misconstrued the idea and thought that only trivial games would have the property (I argued this with Martin and another finance student named Mia), but then Penna introduced “congestion games”. A congestion game is a collection of resources; players have to choose a subset of the set of resources satisfying some condition–a subset of all roads that take the player from origin to destination, say. Then, each resource chosen incurs a delay on each player that chose it, which is an increasing function of the number of players on the resource.
All congestion games are potential games, by a somewhat odd construction–this shows they proceed towards equilibrium in an orderly fashion. Pretty neat!
Then I went to lunch with Martin :) We had a fun conversation about the math of potential games. Found out that his university didn’t even have an exchange setup with the ETH–he argued his way in. What a boss.
Went to an info security seminar that wasn’t starting the first day, so sat and made sure my course registrations were squared away. Then went outside and looked at the first machine learning project. Got hung up trying to figure out how to set up a Linux environment for myself (they didn’t support anyone who tried to do it on a Mac or Windows). I don’t have enough space for a VirtualBox so I think I’m going to just use one of the lab machines. I still don’t understand how the lab machines can keep, say, 10GB of my personal data on a network file system, then load it onto whichever computer I log into. But Daniel convinced me it was doable with fast physical cable connections.
Went back to Game Theory exercises, which were kinda frustrating, because (maybe I just have more background in game theory) figuring out whether to answer or let other people answer was harder than the problems. Except for the last problem, where we had to construct a potential function :)
After that, went to an Apero, or appetizer meetup. Grabbed Daniel on the way. Ran into Aleks, John, Martin, Simon at the Apero and hung around eating tasty meat and cheese.
After the Apero, went back to Culmann, walked with Abhi and played frisbee, then went up to the roof with Juyoung and her buddies Ruth and Jonathan, then made myself some dinner, finished my North Korea book, then got sucked into a game of Bang! with some other people. Decided to take it seriously because I didn’t have the energy to take it sillily; eventually, 2 hours later(!) delivered the fatal Bang! to the Sheriff.
Finally, booked a doctor’s appointment for my cough and my stuffed-up ears, wrote this log, and hit the hay.