I’ve worked in Baar for two months now and I go to work by train. It takes a while longer than to Schwyz as before, but I don’t have to switch trains or busses. That means it’s good for reading. Currently I read “Artificial Intelligence: A Modern Approach” which is accompanying an online curse (ai-class.com from Stanford) that I currently attend. Today I was reading in the chapter about neuronal networks. There it describes an algorithm called “optimal brain damage“. It tries to find an optimal topology for the NN by randomly cutting connections from an initially fully connected NN.While it describes adequately what the algorithm does, it struck me awkwardly when I first read the name.
What are the best names for algorithms you have come across?
After my father suffered a brain bleeding two years ago (he recovered quite well), I read a couple of books on how the brain works and tried to connect that knowledge with the books about artificial intelligence that I’ve been reading for many years. What came to my mind first when I read that, was the “split brain” phenomenon well described by Koch. For extreme cases of epilepsy they cut out parts of the brain or even cut the main connection between the two hemispheres. While the overall effect is positive, the side effects can be quite creepy.
The course (ai-class.com) is actually quite interesting. At first I didn’t know if I’d have enough time besides family and job to do the advanced track, but so far I’m doing good. Last week was the mid-term exam. And I’m already thinking of attending the machine learning class next. I’ve had interest in machine learning since I developed the document type classification and data extraction engines at TCG. Later I used support vector machines and evolutionary algorithms for my free flying prediction system. I found that I learn better with this online course than by just reading books or watching videos. It’s really interactive and still it scales well. More than 135’000 people signed up. I would be interested to know how many are still in. Of course it’s not AS good as attending a real class where you have direct interaction with the profs and peers, but it’s the closest that I experienced so far. And it’s free! Although, if they ask for donations at the end, I surely will.
Looks like 23’000 students completed the mid-term and the average rating was 83%. Â I did 91% hooray! 🙂