Machine Learning (Cornell,UT-Austin,USC,UW-Madison)

[Cute_Jeremy]

I could not make up my mind on which school is the best for machine learning.

I kinda like Cornell because it is a school with a lot of culture, but I am not so sure if its CS is very good.

UT-Austin is famous for its CS program.

Any suggestions?

[mo7aisen]

I understood you are asking about the ranking of these schools in CS. While no ranking is perfect, here's the CS ranking of the USNews (2008).

6. Cornell University.

9. University of Texas Austin

11 UW-Madison

20. USC

I'd rather recommend you to consider the performance of the machine learning group at each school for which Cornell probably wins.

[defmacro]

I could not make up my mind on which school is the best for machine learning.

I kinda like Cornell because it is a school with a lot of culture, but I am not so sure if its CS is very good.

UT-Austin is famous for its CS program.

Any suggestions?

I'd say Cornell, Austin and Madison are all (more or less) equally well known for their CS program, but specifically in Machine Learning I do not know. Ask your professors that do Machine Learning, read papers and talk to professors. At this point, word of mouth is the best way to learn about this kind of stuff.

[bernard]

Cornell used to be very good in computer science, but not anymore. I think it may be due to its location, which cannot attract many good professors. UT-Austin has a reputation in computer science, and is in a much better location. This is just my opinion.

[SingaporeGuy]

just curious, but why did you say that cornell is no longer good in CS?

[tedjj]

I don't see how TX is better than NY

[bernard]

Some of the big names got their PhD in Cornell, and that was a long time ago (like in the 70s). I think UT-Austin is more research intensive than Cornell right now in CS. The reason that Stanford, Berkeley continue to be strong is because of its California location, which can keep professors there for a long time. Even UCSD, USC are rising quickly for this reason. So, if you want to work with world renowend profs, better think location!

[Cute_Jeremy]

Some of the big names got their PhD in Cornell, and that was a long time ago (like in the 70s). I think UT-Austin is more research intensive than Cornell right now in CS. The reason that Stanford, Berkeley continue to be strong is because of its California location, which can keep professors there for a long time. Even UCSD, USC are rising quickly for this reason. So, if you want to work with world renowend profs, better think location!

Thanks bernard, your suggestion has some in sight. I will definitely think it over.

[opsimath]

I have to disagree with bernard here. As someone with a fellowship offer at UT who was rejected from Cornell, I would love to be able to say that UT is "better" than Cornell, but overall, I think Cornell is stronger than UT. You can find some areas where UT has more top faculty members (like comp arch), but there are more areas (like theory) where Cornell has a stronger program.

I don't see how TX is better than NY

Ecological fallacy. This is like saying Yahatanishi must be denser than Manhattan because, gee, I don't see how the US is denser than Japan.

[bernard]

Selectivity does not indicate strength of department.

And have a look at this:

http://www.scribd.com/doc/8508917/World ... etrics2008

[opsimath]

Whoa! The data in that link floored me. I've revised my opinion from being sure that Cornell is "better" to being unsure about the matter.

For the sake of argument, here's one way in which their metric is flawed. Consider a simple model where you only have two broad areas, systems and theory. IMO, it's indisputable that Cornell is much stronger than UT in theory, and that UT is somewhat stronger than Cornell in systems. Just looking at that, I think it's reasonable to say that Cornell is "better" overall. But, many more systems papers are published than theory papers, so if you use a metric based on papers and citations like they do, UT will dominate, and I just don't that's meaningful. Moreover, even if that's how you decide to define quality, who cares? If you're a theory person, Cornell is the better* place to be regardless of the strength of UT's systems area, and visa versa.

Just to clarify, I wasn't making an argument based on selectively. I don't think there's a total order (or even a rank order, or, for that matter, even a partial order) on students across schools. I suspect it's statistically harder to get a fellowship at UT than it is to "only" get admitted at Cornell, and that UT and Cornell simply use different criteria. I just wanted to point out that even though I should be biased towards UT, I consider Cornell the stronger department. I probably shouldn't try to convince people that the department that I'll probably end up in is weaker than a department I've been rejected from though, so I'll stop here

*and of course this depends on what area you want to work on within theory. If you want to work with formal logic in ACL2, UT is clearly the better place to be, and it doesn't really matter that Cornell is arguably the top place in the entire world to do algorithms.

[husker207]

Also, the rankings seem heavily biased towards institutions producing a larger number of papers. Larger faculty size usually implies a larger number of papers. No wonder all the big public schools texas, uiuc,michigan,minnesotta feature in their rankings while strong smaller programs Penn,Brown,Duke,Rice,Caltech don't.

[bernard]

Let me say first that the placement of schools in the ranking does not really matter. The good thing about this ranking is that it is based on objective factors (citations, total papers), and no subjective factors (peer-review) were used. But I also agree with some of you that it does emphasize too much on citations and not putting enough weight on citations/paper (for smaller programs).

The ranking does tell you that Cornell is not on the list, which means it is not a hot place to do research (at least overall speaking). The ranking uses more recent papers (1996-2006) for its analysis, so no ancient classic papers were used. :wink: Perhaps Cornell was more active in the old days. However, if you believe that Cornell has the right PI for you, choose it by all means. But in ML, I think UT-Austin is slightly better.

Bottom line: do not trust ranking too much. Visit both schools to see where is a better fit, both academically and socially. However, if you want to remain in academia after PhD, go to a place where you are more likely to publish (hence look at places that publish a lot).