CMU vs UC Berkeley Statistics PhD

[statistican]

I am deciding between joining a statistics PhD at CMU vs Berkeley. I am very confused at the moment and wanted to get thoughts from the forum. The current factors in my decision:

Berkeley:

• Considered/ranked one of the top two programs in the country
• More "big-name" faculty (Jordan, Wainwright, Yu)
• Greater emphasis on theoretical stats/probability vs applied stats
• Majority of students seem to have one advisor and you make a commitment early in the program as to your research area

CMU:

• Strong connections to the ML department, which is generally ranked number one overall
• A bit more focus on applied stats
• Younger faculty who may be easier to access and get involved with research-wise
• Easier to work with multiple advisors and move across research areas

My research interests are in foundational topics in machine learning (robustness, optimization, etc.) and possibly time series. I want to focus on developing new methodology and honing my practical skills by using the techniques on interesting real world datasets.

Is there anything I am missing here? The main issue I am struggling with is just the fact that Berkeley is consistently ranked higher, but CMU generally seems like a more flexible program.

[Spaghettini Plot]

I think you really can't go wrong.  Those are two of the best programs and either will be a fantastic choice.  One thought is if you know there is one person that you absolutely want to work with at one of these schools, then go to that one.  However, if that is not the case, one method to help decide is to look at which program has more professors you would potentially want to work with.  I don't think that you can fully get an idea of what working with a particular professor may be like before you actually are there, and having several options is very important.  Also as of late, several big companies have a tendency to poach faculty so this is another reason to make sure that you have several people you would be happy working with. 

One other thing to consider is the cost of living is much lower in Pittsburgh than the Bay area.  You will be much more comfortable on your stipend at CMU if that is an issue. 

 

As for the main issue you are struggling with, I wouldn't worry too much about Berkeley being consistently ranked higher, as that difference is only slight.   

 

Best of luck on your decision!

 

[bayessays]

Agreed that these departments are both good enough that you shouldn't choose Berkeley just because of ranking - CMU has some amazing top people.  I guess Jordan is in sort of another league for ML in terms of how widely cited he is, but I imagine there is quite a bit of competition to work with him and he has lots of students/post-docs, so I'm not sure making a decision based on one professor is a good idea. But CMU has people like Wasserman, Kass, Shalizi who are just as widely known as any prof at Berkeley.  

Agreed with above about cost of living. CMU has a generous stipend that will allow you to live on your own in Pittsburgh, but you will struggle to get by on the stipend with roommates in Berkeley.

[bayessays]

I just checked their webpages, and Jordan currently has 33 students and post-docs which is more than the entire stat ML group at CMU with like a dozen good professors has.

[statsnow]

I have familiarity with Berkeley.  Berkeley is consistently rated number 1 or 2 in stats, math and CS.

Berkeley stats is much bigger than CMU.  There is an extremely close tie between EECS and stats at Berkeley. Many of the stats professors have joint appointments with EECS.  Berkeley also has close ties with Stanford, Harvard and Silicon Valley.  It is extremely easy to get a summer internship in Silicon Valley because of its close ties.  Those internships can pay around 10k per month.

Berkeley has lots of young faculty that are eager to work with new students .  Look at Ding, Feller, Fithian, Pimental and Steinhardt.  In addition there are a number of postdocs at Berkeley.  Berkeley has almost 4 times as many postdocs as CMU.

Most students at Berkeley have two or more advisers.  There is lots of flexibility in terms of  selecting your advisers.  They do not all have to come from the stats department. Many students wait a couple of years to pick their advisers.  There is also no qualifying exams at Berkeley. There are no prescribed classes that need to be taken .  Many students easily move across research areas.  And it fact it is strongly encouraged at Berkeley

In terms of ML Berkeley is consistently rated at the top along with CMU.  You may want to look at the Yu group or BAIR.  Berkeley Artificial Intelligence Research has 30 professors associated with it and over 200 grad students and post docs. BAIR is doing cutting edge research in a multitude of areas

In terms of flexibility I dont think there are departments that come even close to Berkeley.  

[statsnow]

To clarify  what bayessays just posted.  Jordan is not the only professor that does ML at Berkeley.  There are probably at least 30 professors and countless more postdocs involved in some form of ML at Berkeley.  

[bayessays]

Yes, I did not mean to imply that. It sounded from the post like OP was leaning towards Berkeley because of 1 or 2 big name professors and I just wanted to express my thought that going to Berkeley because of MJ alone might not be a prudent idea.  Both are great schools and OP can't go wrong. Thank you for clarifying and offering your very helpful knowledge and perspective!

[captivatingCA]

First off, congrats on your acceptances! I just want to add my two cents from my perception of these departments. I'm just an applicant and haven't had extensive experience with these departments, but I can share how things seem to me from talking to grad students and professors.

In general, these departments are more alike than different. statsnow mentioned a lot of positives to Berkeley, and I think CMU shares a good chunk of them. CMU is really flexible in their requirements, and all of the courses are pass/fail. At the virtual visit for CMU, it was mentioned that the funding structure gives students a lot of freedom in advising. So at both Berkeley and CMU you could theoretically have different advisors for all of the projects that make up your thesis. They both have great connections to industry and ML groups. 

That being said, there are differences. To me, it seems that the barrier between students and faculty is lower at CMU. That varies a lot between professors though, and I don't think the difference is substantial. CMU also has a few joint PhD programs, but I'm not sure if your experience will be substantially different if you're in a joint program. Berkeley has more high-ranked departments outside of stats, so there could be more potential collaborators. CMU is smaller, but there's no shortage of collaborators, especially since Pitt is down the street. Berkeley's location is an advantage due to the proximity to other big universities and Silicon Valley. Berkeley seem more theoretical (CMU is pretty applied), but there's a lot of applied people there too. In terms of coursework, CMU has the ADA project. For the most part, the differences aren't that vast.

The big differences between the two will be lifestyle. Given that CMU and Berkeley share a lot in terms of academics, I think it would be easier (maybe even better) to focus on the kind of life you want to live. UC Berkeley and CMU are two very different universities in two very different cities. Since you'll be there for five years, I think you'll maximize your productivity by choosing the place you could live best. 

If it's still difficult at that point, just follow your gut, there's no wrong answer!

[insert_name_here]

I'm a Berkeley grad, but I won't reiterate the pros that have already been mentioned (size/quality of ML profs/postdocs, overall university quality, higher ranking). A few points:

You're worried that "you make a commitment early in the program as to your research area". Functionally, you're expected to find an advisor by the end of your second year (some people do take longer, but not recommended). Idk how CMU works, but 2 years is typically enough to explore. Some people commit as early as first semester if they want.

Berkeley is more expensive than Pittsburgh, but it's also an otherwise awesome place to live. Sunny, not too hot/cold, beautiful nature, spitting distance from SF, wine country, Yosemite, Big Sur. Plus fresh produce. I always hear that Pittsburgh is "nicer than you think", but still...