Unsupervised Learning
by rrusczyk, Jun 26, 2009, 1:48 AM
An AoPS parent just sent me this article on "unsupervised learning".
One key excerpt struck home:
This is how I used math contest to train myself to solve problems. It's a fantastic way to learn, but only if you have good data to learn from. The trouble is, we humans can't process all available data, or even much of it, the way computers can. We humans need an effective filter to present the unsupervised learner with the best set of data to learn from. That's what math contests did for me, and what we're trying to do with AoPS now. (It also might be a future of Alcumus...)
One key excerpt struck home:
Quote:
But in unsupervised learning the student is given, say, the complete set of multiplication exams for the last 20 years--some problems with the right answers, some with the wrong ones--and it is up to the student to make sense of it all.
This is how I used math contest to train myself to solve problems. It's a fantastic way to learn, but only if you have good data to learn from. The trouble is, we humans can't process all available data, or even much of it, the way computers can. We humans need an effective filter to present the unsupervised learner with the best set of data to learn from. That's what math contests did for me, and what we're trying to do with AoPS now. (It also might be a future of Alcumus...)
February 2012
January 2012