MOOC^3

Hey all - in case you're curious at what's been occupying my time recently: www.coursera.org
In particular, while in Zurich I've been filling a bunch of my spare evenings and weekends with some online courses. While I spend most of my day doing computer-science-y stuff (or, I guess 'software engineering' to be more exact), I feel the amount of new stuff that I'm learning now is pretty limited and there's only so much coding I can do before wanting to learn something else... so while my first was a programming course (Scala), the remainder have been a mixture of economics (Economics for Scientists, Competitive Strategy, Advanced Game Theory), med (Computational Neuroscience, Diabetes: Diagnosis, Treatment and Opportunities) and psychology (Beginner's Guide to Irrational Behaviour, Social Psychology, Moralities of Everyday Life).
In particular though, three of those courses mentioned above (Comp Neuro, Game Theory, Moralities) were all offered at the same time at the start of year, when Katie was away and my other commitments (i.e. travel) were low. My plan was to sign up for all of them, and when the workload increased, drop the least interesting one (or two, given I'd only done one at a time beforehand). To cut a long story short, they turned out to be all worth completing, well presented and interesting, and the relatively low workload - a few hours per week for each (thank you, variable speed video playback!) and game theory only lasting four weeks - meant that I finished each. I'd say the Moralities one was the most interesting (probably due to being furthest from comp. sci) and the best taught course I've had so far, but also probably the least challenging out of the three, which I guess suited me as I'm more familiar with the pure maths of game theory, or the coding/modelling for the neuroscience course. Hilights from each?

Game Theory

The Revenue Equivalence theorem - roughly speaking, when distributing goods to rational, risk-neutral people, then given two distribution mechanisms (e.g. different types of auctions), if they both distribute goods in the same way given the same inputs (e.g. bids), and don't distribute anything to someone who values the goods as zero, then the expected revenue of both systems is equivalent. This is extremely general, but means that e.g. any auction of one item that gives the item to the highest bidder will have the same expected revenue as a second-price auction (i.e. the winner pays the amount of the second highest bid). Which is why most bidding systems (e.g. auctions like ebay) are essentially this. In a close second place, the Vickrey-Clarke-Groves auction is a nice auction that both maximises global utility ('socially optimal'), but also makes it optimal for people to value things at their true personal utility (and has a nice proof).

Computational Neuroscience

The most 'woah' style moment came right at the start - one of the early pracs included data from the firing rates of cricket neurons which were attached to leg hairs which sensed wind direction. It had been discovered that these four neurons fired related to the angle of the wind relative to 45/135/225/315 degrees respectively, and so by applying some algorithms, you could calculate the direction the wind was blowing on the cricket in the lab at the time. Which is pretty trippy once you think about it! Other neat things included examples of how simple neuron models turn out to implement quite complex algorithms (calculus, stats, even principal component analysis / k-means clustering), and a somewhat freaky guest lecture where scientists were able to train a monkey to move their hand by sending neural signals through an external computer, while the normal link (brain -> spine -> arm muscle nerves) had been cut. Future implications for this are both impressive (e.g. fixing people with spinal damage) but scary (digitally 'enhanced' neural processing).

Moralities of Everyday life

Not really any wow! moments here, but still interesting - I think the experiment that I remembered the most, other than ones I knew already, was the following progression. First, you start with the Ulimatum Game: given $X (in the order of $10), you get to offer some to someone in another room, who can't see you and can't talk back. If they accept your offer, they get it, and you keep the rest. If they reject it, you both get nothing. How much do you offer? (average is about 50%). Then you progress to the Dictator Game: same as above, but they can't reject it - you get to keep everything not offered. How much do you offer in this case, where there's no threat of retaliation by the other party? (average is about 20-30%). Finally, it progressed to a variant of the dictator game: before playing, you're given the opportunity to walk away with 90% of the money - i.e. in the $10 example, you can either take $9, or play the dictator game (and get $0 to $10). Without external factors, it is irrational to take the $9 - either you don't care about the other person, in which case you are better of playing the dictator game and keeping all $10, or you do care about them, so you should play and at least offer them $1. That said, many people (me included!) would take the $9 option, and the theory is that it's worth the difference just to not have to decide in the dictator game.

That's all for this update - hopefully travel starting again soon (London and Australia booked!). In the meantime, I have started another course (Ethics, by the Australian Peter Singer), and am otherwise now keeping myself busy with this awesome game (2048, pictured above) plus finally getting Gran Turismo 6 as a reward for finishing the three courses - custom music, mount panorama plus the return of the red bull concept cars means it'll occupy my spare time for a while!

Comments

Popular posts from this blog

Perceiving Frequencies

Project cutting floor

Unhealty healthcare