September 30, 2006

Mathematics of Generalization

[Learning_ Books_] Mathematics of Generalization was one of those books that had been waiting on my shelf for quite some time. Partly because I was intimidated by the possible difficulty of mathematics. It turns out the articles are very well written and not that intimidating at all. I will post small perls of wisdom I get from the book in the comments of this post.

Related link

1 comment:

Deniz Yuret said...

A learning problem is described as figuring out a mapping between from a set X to a set Y. The interesting observation is even if the set X is non-countable, the mappings induced by any learning algorithm (being programs) are countable. So the learning algorithm cannot access the uncountable infinity of the input space and we don't lose anything by restricting ourselves to a discussion of countable spaces.