Re:Research Blog: Machine Learning Book for Students and Res...
by imurray on Wednesday August 22, 2012 @12:56PM
(#4418305)
Aside: I find it ridiculous that I can't read a blog post, a little bit of text and an image, without enabling javascript on two domains. I know that a bunch of you will think only masochists use NoScript, but the page is slow to appear and needlessly complicated. If there's anyone at Google who could help your company set a better example, please nudge the right people.
Re:Research Blog: Machine Learning Book for Students and Res...
by pasbesoin on Wednesday August 22, 2012 @03:30PM
(#4419129)
And this one is not in Google's cache. I've taken to reading the cached version of many blogspot pages, for this reason -- I have a browser extension that makes that just a couple of clicks to accomplish. Maybe Google has caught on to this.
EDIT: My mistake; I just had to strip off the bazillion query parameters.
Re:Research Blog: Machine Learning Book for Students and Res...
by iskander on Wednesday August 22, 2012 @12:21PM
(#4418133)
I haven't seen this book before but I did take Prof. Mohri's class a few years ago. His material is great but be warned that is's biased heavily towards theory (proofs and bounds instead of the more common/practical "rules of thumb"). You can get a lot done using machine learning tools having only a superficial familiarity with VC dimension, complexity bounds, etc... If however, you want to get deeper insight into the algorithms you're using or develop new algorithms yourself then Prof Mohri's of rigor is very useful.
Re:Research Blog: Machine Learning Book for Students and Res...
by achompas on Wednesday August 22, 2012 @05:19PM
(#4419640)
I took Prof. Mohri's class this spring, and we actually used a draft version of this book. I agree with you that it's very theoretical, but Bishop, Mitchell, and the other ML resources don't address learning theory. VC-dimension isn't for everyone, but if you need to know about it, this is the book you want to use.
There's also the added bonus of this text coming from Prof. Mohri et al.'s experiences at Google, so there's a lot of discussion about online algorithms and ranking (which you don't see in many other places).
Re:Research Blog: Machine Learning Book for Students and Res...
by exg on Wednesday August 22, 2012 @01:43PM
(#4418554)
The Elements of Statistical Learning, by T. Hastie, R. Tibshirani and J. Friedman [1] is also a very good one. Plus, the book is freely available on the authors' website.
Re:Research Blog: Machine Learning Book for Students and Res...
by achompas on Wednesday August 22, 2012 @06:07PM
(#4419880)
So this is a common misconception about the text (and Prof. Mohri's NYU class). In this case, "foundations" does not mean this is an introductory course.
Rather, the class and text provide mathematical foundations for understanding the error bounds and growth complexity of various learning algorithms. So you'll be workin with convex optimization, reproducing kernel Hilbert spaces, and Rademacher complexity--definitely not "introductory" in the least!
It's a completely different beast from Mitchell, Bishop, or EoSL (which I'm studying right now!), so I'm not sure comparisons are valid. It also fills a prominent gap in the ideas reviewed by the popular ML texts.
Re:Research Blog: Machine Learning Book for Students and Res...
by andyjohnson0 on Wednesday August 22, 2012 @12:19PM
(#4418126)
One of the authors (Mohri) teaches a course called Foundation of Machine learning. Details are here[1] and may give an indication of the books' contents.
It's the last class he taught before the book was published. I was in the class, and can confirm the lecture outline closely matches the book's contents.
EDIT: My mistake; I just had to strip off the bazillion query parameters.
Google cache of the post:
http://webcache.googleusercontent.com/search?q=cache:http%3A...
There's also the added bonus of this text coming from Prof. Mohri et al.'s experiences at Google, so there's a lot of discussion about online algorithms and ranking (which you don't see in many other places).
The Bishop book is the most popular though: http://www.amazon.com/gp/product/0387310738/ref=pd_rvi_gw_2/...
[1] http://www-stat.stanford.edu/~tibs/ElemStatLearn/
Rather, the class and text provide mathematical foundations for understanding the error bounds and growth complexity of various learning algorithms. So you'll be workin with convex optimization, reproducing kernel Hilbert spaces, and Rademacher complexity--definitely not "introductory" in the least!
It's a completely different beast from Mitchell, Bishop, or EoSL (which I'm studying right now!), so I'm not sure comparisons are valid. It also fills a prominent gap in the ideas reviewed by the popular ML texts.
[1] http://www.cs.nyu.edu/~mohri/ml10/
http://www.cs.nyu.edu/~mohri/ml12/
It's the last class he taught before the book was published. I was in the class, and can confirm the lecture outline closely matches the book's contents.
http://www.nytimes.com/2012/08/12/business/how-big-data-beca...