Friday, November 14, 2008

Learn from the Best

I play ping pong quite a bit and I started to play better after I watched some of the games of the best player at work. I kind of learnt some of the game strategies by observing his game. I applied those in my game and it worked pretty good. So that brought me to my utopian land again. How about having a machine learning algorithm that learns from the "best player" data set instead of a huge training data set. That would be just great - Learn the best strategies and moves from the best player. For eg, We try to get into the best schools so that we can learn from the best professors, rite. Why not adopt this idea to machine learning ? I searched a little bit but couldnt really find a place where this idea is being used. Now the problem is, "How to know whatz the 'best player' data" ? well, I thought of having some ranking algorithm that would point us to the "best player" data. Now how will this ranking algo operate or work ?? hm.. thatz a black box for me now.. still need to dig deep and see how to construct it, i guess. If you have any thots, plz ping :)

No comments: