The taste of Vanilla Coke reminds me of working on CS360 (Theory of Computation?) assignments last term in the Math C&D. This is where I am, and what I'm drinking, right now as I work on my CS486 neural networks assignment. The Math building is a weird place at night (actually anytime) what with all the people wearing costumes and their role playing games, and the frosh math girls (??) playing hockey in the hall with a crushed tin can for a puck. I'd estimate that half of the people are working on assignments, here because they have to be, and the other half are just hanging around, here because where else would they go?
Anyway, let me tell you a bit about neural networks, my new pet topic:
There are two ways to view neural networks. The first is as a method of representing functions using networks of of simple arithmetic computing elements which can be constructed from examples. The second, more alluring, viewpoint sees neural networks as a mathematical model for the operation of the brain. The simple arithmetic computing elements correspond to neurons -- the cells that perform information processing -- and the network as a whole corresponds to a collection of interconnected neurons (ie. the brain).
Our assignment is to train a neural net. to recognize bitmaps of the characters 'A' through 'E'. This exact process is used by the post office to sort mail by having computers "read" handwritten postal codes. It seems to me that a CS graduate could get rich by selling a big company software that automates an expensive human labour-intensive part of their business ... But those systems have an error rate of about 1% and right now mine is at no better than 10% so back to work.
And by the way, I talked to the nice people at Net Integration Technologies and they will be happy to have me back next term. So, summer in Montreal. I think it will be nice. <Jealous Voice>But not as nice as it would be to make twice the money in Seattle like everyone else is.</Jealous Voice>
8:45:55 PM 
|
|
