I’ve already blogged once about the first miniproject of my class this term, rankmaniac, and now after a one-week break, it’s time for the second miniproject of the term: pandemaniac.
This is the part of the course that I’ve been looking forward to the most since the beginning of the term. We’ve been prepping the infrastructure for the assignment since the Fall and now, thanks to two impressive undergrads (Max Hirschhorn and Angela Gong), we have a nice system built that will allow the students in the class to seed competing cascades and then watch them play out step by step…but, I’m getting ahead of myself.
The idea of this project is to crown the “pandemaniac,” i.e., the king/queen of pandemics in the class. In brief, what the students will have to do is to write an algorithm that, when given a graph and a number of competing players, will determine which nodes to use to seed a pandemic/epidemic/cascade. Many teams will choose seed nodes on the same graph, and the idea is to choose your seed nodes such that your epidemic takes over the largest section of the graph.
To help make it fun, we’ve divided the assignment into two parts: a regular season and a tournament. During most of the one-week assignment, every night, the students can play randomly-matched games with each other and with the TA teams. This should help polish the algorithms and explore the design space… Then, at the end of the week, we’ll have a whittle-down tournament where each round will narrow the number of teams left by half until we crown the pandemaniac. (Of course, to allow for randomized policies, each round will include multiple graphs, of a variety of sizes.)
As I’ve discussed, I like the motivation provided by tournaments, but I worry about some of the side effects. So, the grade is not tied to the tournament at all — the bulk of the grade comes from a report describing the strategies the teams tried, and most of the rest of the grade comes from beating the TA teams, which use simple strategies. That said — of course, the tournament provides nice motivation for the students to try interesting strategies! …and we’ll then spend the lecture following the miniproject in discussion.
I’ve been really excited about this project because I’m extremely curious about how people will actually compete when trying to create cascades. There have been a number of papers in recent years that try to capture equilibrium strategies in settings where there is competition among cascades, but it is really very difficult to define models that are both believable and tractable in the space. I figure with 15-20 teams of Caltech students playing for a week, we’ll get a good approximation of some equilibrium strategies! (Of course, with the caveat that we’re using a very simplistic model for the spread of a cascade.) In any case, at minimum, the results of the competition should be an interesting data point about what strategies could emerge in a context where people are competing to seed cascades…
If you want to follow along…
Good luck everyone!