Re:New Book from O'Reilly: Programming Collective Intelligence
by snifty on Sunday August 26, 2007 @04:14AM
(#46703)
Just because people are interested in "Web2.0" doesn't mean they're not going to be interested in the kind of stuff in this book.
In my opinion, interest in this sort of material is going to be propelled forward by Web2.0. It's not at all true that web programmers "don't give a shit" about mathematical modelling, etc. It's just that there hasn't been much accessible code for people to look at (much of it tangled in obscure, obtuse, proprietary academic and military labs).
It's true that this book seems to be attempting to cover a lot of topics, but that's probably the point: a decade ago, there was no popular concept, let alone interest, in relational databases. That changed because the street found a use for it, as the saying goes. Anyway, I'll second your Witten and Frank recommendation, cool book.
Huh? PageRank dates from the late 90s. And it's implemented in a compiled language (C++, I think -- not even GC'd).
And it involves elementary linear algebra. Heresy!
Now, as for the book itself. It looks like it tries to cover waaay too much -- support vector machines and other forms of supervised learning, unsupervised learning/clustering, optimization and evolutionary computation, applications to collaborative filtering, along with using particular libraries. There's no way you could cover all that in 358 pages!
I gave a talk on evolutionary computation to a Ruby users group and their jaws literally dropped. Web people don't give a shit about something if it doesn't involve HTTP, a programming language that's currently en vogue, JavaScript, or relational databases (really only MySQL or maybe Postgres).
If you're interested in an introductory data mining book with a practical focus, may I instead suggest Witten and Frank's "Data Mining: Practical Machine Learning Tools and Techniques"? It uses Weka throughout, which is mature and nice.