Its the time of year when Seniors who want to go
to Graduate School should be pondering WHERE to go.
There are other issues which I will address in later
posts, but
Today's topic is WHERE TO APPLY?
I would like YOU (the readers,
and time magazines MAN OF THE YEAR for 2006) to comment on:
-
IF a student wants to do COMPLEXITY THEORY where should she go?
-
IF a student wants to do COMBINATORICS (in a math dept) where
should he go?
-
IF a student wants to do XXX (in a YYY dept) where
should ZZZ go?
Note that there may be lesser-well known schools (e.g., not
on someone's arbitrary top 10 ranking) that are really
good in these topics, and they should not be overlooked.
IF a student wants to do XXX (in a YYY dept) where should ZZZ go?
ReplyDeleteMaybe AAA or WWW :-)
You make it sound like a student is free to choose to go wherever she likes. In reality this is usually not the case (i.e. they may not get admitted).
ReplyDeleteAnonymous 2: where should a student even apply?
ReplyDeletehttp://weblog.fortnow.com/2006/03/choosing-graduate-schools.html
ReplyDeleteGo to UC Berkeley, MIT, or Princeton.
ReplyDeleteGo to UC Berkeley, MIT, or Princeton.
ReplyDeleteThats good advice for complexity theory. For some other fields?
Crypto: MIT, Princeton, Stanford
Learning Theory: CMU, Harvard, Penn, Columbia
Game Theory: Berkeley, CMU, Stanford, Cornell
Note that there may be lesser-well known schools (e.g., not on someone's arbitrary top 10 ranking) that are really good in these topics, and they should not be overlooked.
ReplyDeleteThis brings up the whole can of worms of how much overall program quality matters. If there's a great advisor at a less than great school, and if you are really sure you want to work with that advisor, then how would this compare to attending a top graduate program? There are a lot of issues here (interactions with other students, reputation effects that may or may not be deserved, etc.), and I doubt a detailed discussion will prove illuminating. However, I'd be very curious to hear some answers to the following slightly more specific version of the question:
Who are some great advisors who aren't at schools widely considered to be in (or near) the top 10?
For example, if you want to do complexity or crypto but aren't likely to be admitted to Berkeley, MIT, Princeton, or Stanford, then what are some other good options?
Here's a concrete but out of date example. Doron Zeilberger is currently at Rutgers, which is a great place to do combinatorics. For many years, he was at Temple, which has a solid graduate program but is not remotely competitive with Rutgers. Back when he was at Temple, students who were admitted there but not Rutgers had a great opportunity. What similar opportunities are out there now?
Crypto: Stanford??? Dan Boneh is there, but still... A place like UCLA is a much better bet.
ReplyDeleteSome self-explanatory programs (only one is a shameless plug!)
ReplyDelete"algorithms, complexity & optimization" @ cmu
"algorithms, combinatorics & optimization" @ gatech
"combinatorics & optimization" @ waterloo
distributed systems
ReplyDeleteCornell, MIT, ETHZ
What about schools outside the US?
ReplyDeleteI'd say that Weizmann Institute is pretty good for Crypto/Complexity crowd [with, alphabetically and among others, Goldreich, Naor, Reingold and Shamir].
Of course, you'd have to aim for and Msc before your phd.
Weizmann is great,
ReplyDeleteThey also got Goldwasser (part-time), Feige and Raz
Harvard, Yale, Courant have great theoreticians and amazing math departments.
ReplyDeleteYale is small but strong (Feigenbaum, Spielman). Harvard has Mitzenmacher, Rabin, Vadhan, Valiant. Courant has Cole, Dodis, Khot, Naor, Pach, Spencer. All great options.
ReplyDeleteSome self-explanatory programs (only one is a shameless plug!)
ReplyDelete"algorithms, complexity & optimization" @ cmu
"algorithms, combinatorics & optimization" @ gatech
"combinatorics & optimization" @ waterloo
Just a correction: The ACO program at CMU is also "algorithms, combinatorics, and optimization," not "complexity," not that it really matters. Incidentally, Is the ACO program stronger at CMU or GaTech? Seems like Tech is stronger combinatorics and optimization, but CMU is perhaps slightly better in CS.
In terms of students, the CMU ACO program is the much stronger of the two, without question. In terms of faculty, it is up for debate, but I think CMU still gets the edge.
ReplyDeleteI'm surprised nobody has mentioned this yet, but for Crypto: UCSD
ReplyDeleteSome other very good options for Theory:
ReplyDeleteUniversity of Washington, Seattle
UCSD
University of Chicago
University of Toronto
University of Waterloo
UCSD is also very good for Learning Theory, with Sanjoy Dasgupta, and Yoav Freund, among others.
ReplyDeleteColumbia has a strong theory group with Yanakakis, Stein, Malkin, Servidio.
ReplyDeleteFor any specific subarea of theory there are many very good professors at not a top 10 school. Someone asked for specific examples of such opportunities, so here is one: Penn state has Adam Smith and Sofya Raskhodnikova.
Weizmann also has Irit Dinur now.
ReplyDeleteNo one has mentioned one of the OBVIOUS ways to choose among several acceptable options for grad school. If you're going to spend 4-6 years of your life somewhere you should do it in a town/city that you would otherwise want to live in.
ReplyDeleteUniversity of Rochester is the obvious choice for computational politics.
ReplyDeleteThere are too many anonymous comments in this thread. Last time the topic was discussed I considered putting an anonymous comment that would list a group of schools including mine, but decided against it and just waited (in vain) for my school's name to appear. Now I get a second chance!
ReplyDeleteIn fact I could post several comments anonymously with several different lists which would happen to all include my school. Is anyone out there using such advertising strategy? Among the schools cited so far, is there any that looks ever so slightly out of place?
If there's a great advisor at a less than great school, and if you are really sure you want to work with that advisor, then how would this compare to attending a top graduate program?
ReplyDeleteIf your goal is an academic job, especially in the current job market, then you essentially MUST go to a top school. For those who doubt this, please try to name a theorist hired in the past 5 years at a top-50 school who graduated from a school below the top 10. Off the top of my head, I can't think of any (though I'm sure there must be some).
You will also in general be better off at a top school because of the quality and quantity of theory students you will be surrounded with. Having the best advisor in the world will not compensate for being one of only 2-3 theory grad students in a department.
Claire, it's funny how your expressed exactly what I felt :). Since the posts are anonymous, it's hard to take them seriously since it could be simply shameless self-promotion, done in an anonymous manner. (Of course, since I could claim to be anybody I want, putting a name does not solve the credibility issue, but at least it allows for what we call a "disavowal" protocol in cryptography; here, hopefully I proved my identity :).)
ReplyDeleteOn this note, I was very glad that NYU was mentioned, and I can swear that even the post with my name was not done by me :). Speaking on NYU, though, I will non-anonymously add that I think we've done a great job in the past few years to hire top people like Assaf Naor and Subhash Khot, and I'm really looking forward to working with them.
Connecting this with the previous post, I agree that it is a big disadvantage if you go to a school with a small theory core, even if it has an amazing singleton adviser. Indeed, the chance of getting a good academic job is zero in this case (of course, among other reasons this is correlated with the quality of the program and the size of the theory group, but still the name matters A LOT).
Thus, I hope that, with the recent hires, NYU will rapidly go up in the rankings of the Theory programs, and we'll be able to get top students, who in turn will have great academic job prospects. And, btw, did I mention we are in the center of the best city in the world :) (in addition to now also having a strong Theory program)?
Among the schools cited so far, is there any that looks ever so slightly out of place?
ReplyDeletehttp://nt.cra.org/scripts/rankcs.pl?TOP=108&DIRECTION=DESC&ORDER=PUBTOFAC&FIELDS=CITEFAC
Penn State?
I see in that ranking MIT narrowly loses to University of Central Florida.
ReplyDeleteYevginy's post brings up some issues - if the name of your PhD granting institution matters so much, where does it fall in the criteria for making a new hire?
ReplyDeleteThe usual criteria that one thinks of when making a hire are "has strong results", "is productive", "works in an interesting area", "is likely to get funding", "can teach some classes that we need to cover", "pleasant to work with", etc
But at which point does "this person got their PhD from a Program X and therefore must be better the other person with a PhD from Program Y" become a factor? Is it just a tie-breaker, or does program reputation actually take more weight than some of the other criteria?
Or does program reputation actually enter more as a social network -- "we have several faculty from Program X, and they are on good terms with the gang back at program X, so let's be safe and chummy and take another program X graduate" ?
...if the name of your PhD granting institution matters so much, where does it fall in the criteria for making a new hire?
ReplyDeleteBased on my observations over the past 8 years and experiences with my own students, the school where you get your PhD makes a huge difference. Unless you are a superstar, you won't even get the interviews unless you come from a top-ranked school.
Do the research yourself: look at the top 25 schools, and see who they brought in for interviews over the past year or two (you can usually determine this information by looking at their seminar schedule). See how many of these candidates came from sub top-8 schools
Okay, so that is a description of the situation. I didn't really contest YD's points.
ReplyDeleteBut it doesn't really explain why things are that way.
Is it just that the people from those top eight schools are so much stronger than the competition? That they are so much more numerous?
Or do the search committee members actually assign a non-zero weight to what school you went to?
Surely some people who read this blog have served on search committees and can provide some light on this matter!
Ok Claire, here we go: Brown is the obvious choice if one wants to do interactive computation. Happy?
ReplyDeleteOr do the search committee members actually assign a non-zero weight to what school you went to?
ReplyDeleteNot necessarily explicitly, but when you are looking through a stack (ok a web database) with hundreds of applicants, you tend to only look closely at applicants whose institutions stand out. (Note that PhD institution is immediately apparent, while determining number of publications requires actually looking at the person's CV, etc.)
TCS does not reside in a vacuum. Your maturity as a computer scientist increases if you appreciate all areas of CS as much as you do TCS. In that sense, a top-tier department offers you a much better exposure of all areas of CS than a lower-ranked school. And it is my hypothesis that that this is what causes hiring committees to give more value to students from name-brand schools.
ReplyDeleteI am rather impressed with Georgia Tech's ACO program. It breaks down a lot of the walls between math/cs/engineering. IMHO we would be all better off if grad schools only made students take 4 classes in their home department and let the committee decide the rest.
ReplyDeleteIs it just that the people from those top eight schools are so much stronger than the competition? That they are so much more numerous?
ReplyDeleteOr do the search committee members actually assign a non-zero weight to what school you went to?
Here's my experience (on some search committees but for less than a decade):
Where you went to school isn't a big factor. Going to a top school gives you a tiny boost and it keeps your application from being overlooked, but it won't help you much. What's absolutely crucial is letters of recommendation. You've got to have letters from people who are well known and have long track records of evaluating lots of great graduate students.
A letter saying you are one of the top students from Berkeley or MIT pretty much guarantees that you are excellent, because you are being compared to lots of known quantities. A letter saying you are the best student in decades from the 50th-ranked CS graduate program doesn't mean much, because who knows who the comparison group is? You might be amazing, but this letter will always lose out against the ones the students from the top schools get.
So I believe it is possible in principle to get a great job from a not so great grad school, but it's a lot harder. You have to do great research, do it early enough to get noticed widely, have lots of contact with faculty members from top schools, and get great letters from them. If all your recommenders are from your own school, you'll get a job but it won't be at a top school.
Another important effect to keep in mind is that every school produces far more graduate students than it ever hires, by an order of magnitude or more. This guarantees that most students will get jobs at much lower-ranked schools than the ones they studied at.
If your goal is an academic job, especially in the current job market, then you essentially MUST go to a top school.
ReplyDeleteThe data doesn't back this up.
For those who doubt this, please try to name a theorist hired in the past 5 years at a top-50 school who graduated from a school below the top 10. Off the top of my head, I can't think of any (though I'm sure there must be some).
There aren't enough theory positions in five years to make this a representative sample, particularly considering that, on the average, the best students go to the top 5 schools.
Cristian Estan from Wisconsin-Madison collected data on that.
As you can see there are people hired across the board, and more or less in the proportion one would expect according to relative strengths.
Does it help to go to a good school? Yes, but as a previous poster commented only in a secondary minor way. It vastly overstates the case to say that it makes a "huge difference" and that students "MUST go to a top school".
Take the best offer(*) you get and if that doesn't include the top 5, this in now way means that you are out of the top 50 market for theory positions.
(*) best offer you get is a combination of prestige, strength of school in your area of interest, financial support offered, city, two body problem considerations, etc.
I'd have to say I think people are mixing up correlation and causation here.
ReplyDeleteYou don't need to go to a top school to get a good job. But obviously good people tend to select the top schools. And obviously there is some advantage to going to a top place -- you're in contact with other very bright students and with top faculty who are trying to be at the forefront of their field!
There can be some disadvantage going to a top place. Being the nth best from MIT in given year for large n is not going to get you noticed -- and similarly you might have been better off at a place where you'd get more attention.
The big thing, as another commenter pointed out, is letters. If you go to a lower ranked institution, get out there and do some summer internships, or start a collaboration with someone outside your school, or something! You want letters from top people to get a top job!
Sorry to say, but the link given in comment 37 seems to show the opposite conclusion.* Below the top 15 schools there are only (at most) 2 graduates from a school in current faculty positions! This includes highly-ranked schools like Rice, Brown, Maryland, etc.
ReplyDelete* In any case, the data are not that useful. It is incomplete (I was missing) and includes all years (not just recent years, where I think the trend has gotten worse).
Michael is certainly right that there is some confusion between causation and correlation. But it doesn't matter. If you go to a good school you will be surrounded by good students and professors, get a better selection of classes, get letters of recommendation from better-known people, and have a better shot at internships. Some of these are "fair" (better students/profs/classes) while some are not (rec. letters counting more, better access to internships, name-brand inflation). But in the end it doesn't matter -- either way, there is a strong motivation for going to the best school you get in to.
ReplyDeleteThe big thing, as another commenter pointed out, is letters... You want letters from top people to get a top job!
ReplyDeleteIt is good to understand why these are important. It isn't a matter of getting the approval of someone from some club.
* The obvious part is that top people are expected to have good judgment about the quality and importance of a candidate's research.
* The other key part is that top people are likely to have first-hand experience of interacting with a large number of top students. Because of this, they will have a much stronger basis for ranking a candidate and for making comparisons. How strong is "best student I've supervised in the last n years" or "best student finishing this year in our department" when the typical student is much weaker than at the top places? It is an extremely competitive environment out there and these kinds of statements help to differentiate someone from the pack.
These days, it is pretty competitive both for students and departments. Almost all of the strongest students will get into several of the top 10-15 or so places in any research area. Because of this, the drop-off in the average quality of graduate students is dramatic after that level. This does not mean that places ranked lower won't get good students, or that students can't do well at lower ranked place but it takes more to stand out.
One thing to think about when choosing departments is not just "How strong are they in X?" but also "How broad is their strength?" Lots of students, even those with prior research, think they want to do X and end up doing Y because their view of the landscape from the undergrad level is somewhat limited.
Below the top 15 schools there are only (at most) 2 graduates from a school in current faculty positions!
ReplyDeleteExactly how many top notch students did you expect a second rank university to place in such programs?
The data shows conclusively that the system has enough flexibility to identify and hire stars that for whatever reason did not graduate from a top school. This falsifies the statement that students "must" go to good schools.
In any case, the data are not that useful. It is incomplete. (I was missing) and includes all years (not just recent years, where I think the trend has gotten worse).
Wrong on all three counts:
1) The data is useful even if it were incomplete, as it shows that people are hired from lower schools.
2) It isn't incomplete rather it was collected in 2005, which is why it doesn't include you.
3) It does not include all years as it describes assistant and associate professors, which generally means last 15 years or so. For the data including all years have a look here.
Lastly, I echo Michael in that you are confusing correlation with causation.
I'm wondering if there's currently a good source for grad school applicants to consult for theory rankings in the U.S. These tend to change around quite a bit, as people move and areas go in and out of fashion.
ReplyDeleteMy personal (necessarily somewhat subjective) list of the top 12 schools would be: Berkeley, MIT, Princeton, CMU, Cornell, Harvard, Georgia Tech, Stanford, Washington, San Diego, Texas, NYU. This is based on faculty strength + how well top theory students have performed. I'm willing to bet this is close to the consensus ranking, modulo some differences in the order and a couple of substitutions at the low end.
bah, rankings and careerism and gamesmanship---- what we should ask is "what should the aspiring graduate student do with those years to best become a solid researcher / educator / computing professional?"
ReplyDeleteSome good schools for crypto not in top 10 overall (not in any order):
ReplyDeleteUCSD
NYU
UC Irvine
UC Davis
U Maryland
Georgia Tech
UCLA
Boston U
Brown
when discussig theory, you can't forget that students at Princeton have easy access to the Institute for Advanced study (Avi, Sasha Razborov, Noga Alon, Russell Impagliazzo, and a horde of excellent sabbatical visitors and postdocs, as well as lots of great people who pass through to give talks)
ReplyDelete45, here are a few you missed (though I don't know what you meant by top 10, one of these may be):
ReplyDeleteCornell
Rutgers
Columbia
Indiana University
University of Connecticut
Johns Hopkins University
This is a depressing thread. Obviously not everyone can get into a top-tier or even a second-tier school. I would find it much more interesting if people mentioned 3rd-tier schools that had worthwhile theory departments. (I would also be curious about exactly what graduates from these schools could look forward to doing, since evidently jobs in academia would be out of the question.)
ReplyDeleteI'm glad to see Princeton's name mentioned, but still do NOT think ranking matters for the job market as much as people claim it to be. Theorists will judge a fellow theorist based only on research results, so if anything a biad towards highly ranked schools will be made by people that are too far removed from the field to judge the results directly or even know who are the right people to ask for references.
ReplyDeleteNow of course such people are also involved in hiring decisions, but the theory job market seems to be evolving in the direction of having an almost "mandatory" postdoc phase (a good thing! having people spend a year or two to focus on research is great for the field). A postdoc hire will most likely be decided only by theorist, so your school name on its own will have very little or no effect.
That said, many of the highly ranked schools are great places for a number of reasons and it's good to apply there (in particular, please do consider applying to us.. :) ). But unfortunately not anyone can get in, and every year we are forced not to admit extremely good students. Luckily in theory it's definitely possible to find very good people to work with in a number of places.
Non-self advertisement: By the way, perhaps some people don't know that the language of instruction in the Weizmann Institute of Science (where I went to) is English and they are very welcoming to foreign (i.e., non-Israeli) students. You might have to do an M.Sc first, but you can consider this as just another name for the first 2 years in a Ph.D program in the US.
That said, many of the highly ranked schools are great places for a number of reasons and it's good to apply there.
ReplyDeleteExaclty, but not so much because of the name recognition. They are great places because they have, on the average, the best researchers, the best students and the most funding. Yet if you failed to be accepted at those you can still make it to the top 5 schools as a professor, let alone the top 50.
Boaz-
ReplyDeleteA requried postdoctoral period is not a sign of strength in the community - it is a sign that there are more graduates than positions. Hiring departments can now expect a few years of postdoctoral experience before making a TT hire. If there were more positions (or fewer graduates) they would not be able to have such expectations.
Furthermore, while the postdoctoral period can be fun (you do nothing but research), it is an enormous head-ache for people with two-body problems or families.
Ultimately, such hassles will drive people towards disciplines and careers that allow a more balanced life. By driving people away, such requirements have some cost to the field (although an individual postdoc may be wonderful for the development of an individual researcher).
Another problem with postdocs is ... they are almost impossible to find. NSF gives so little money that one is luck to support his/her students. So the necessity of a postdoc is a very negative sign, since the prospect of finding one is almost as hard as finding an academic job.
ReplyDeleteFinally, regarding finding a good academic job from outside-top-5 institution. I agree with the comments that this is not due to some mafia sitting at hiring committees and snubbing people outside of top 5 places. It's just that even if you are really good and outside of top 5 (which, to be fair, does not happen that often, but does happen), then you are still at a huge disadvantage: (1) there are fewer opportunities to grow within your department (take classes, attend good talks, etc.); (2) it's much harder to meet famous people, (3) even harder to impress famous people, (4) your adviser is less likely to "push" for you behind the curtain (with success). For (3), you must first get some great research accepted at top conference, and then be social enough to make contact during the limited time at the conference.
So it's not that the society is corrupt and snobby, but a combination of factors.
Surely it's a factor that universities would like to say: 30 % of the faculty at our school are from MIT, another 30% from Berkeley... When the focus is on statistics, individual talents and aptitudes get short shrift.
ReplyDeleteAnon 51- Perhaps I should rephrase, I don't think making postdocs mandatory is a good thing. But I do think making more postdoc positions available and having more people do a postdoc is a good thing. Most (though not all) people I know enjoyed their postdoc very much, even when it involved uprooting their spouses and children and moving to a foreign country for a year or two.
ReplyDeleteIndeed, the biggest problem with postdocs is (as Anon 52 said) that we don't have enough of them. If we had more positions in more locations, then solving 2-body problems would also be easier. Now it's not easy to find funding for postdocs since they cost about twice as much as students, but I think all of us should try to do our best to create more such positions.
I am not sure why NSF CISE does not have a postdoc program per se, unlike most other disciplines (though a number of good CS theory people do get NSF Math Sciences postdocs). OTH, there are probably more theory postdoc positions available today than there have ever been, when you add up research labs and institutes along with universities. The fact that the increased number reflects strong interest by research labs and institutes is a very positive thing for the field.
ReplyDeletewhat about machine learning?
ReplyDeleteat 22:
ReplyDeleteTherefore, do NOT go to Chicago.
Good advice in general, actually.
As a current PhD student in computer science (though not in TCS per se), I do believe I have a few words to say about this; though not about schools specifically.
ReplyDeleteFirst of all, as many people pointed out, it's not like you choose grad school, you just try to choose where to apply. I am sure most of the people commenting here were excellent students like Prof. Barak or Prof. Mathieu; however I think for them the choice is no problem, they just look at the best and very probably they get in.
The major hurdle, I would say, is awaited by just the lower class: very good students (clever, eager to learn, good background). Such students are a lot, and if you ask me, at this point, the odds of getting into a top-50 school is dependent on pure luck + of course the obvious variables.
At first sight, I really found it very strange that an academic pursuit is dependent on such luck, but now I find it normal; just because the number of parameters is just too many.
My advice from here on are for students abroad thinking of applying to grad school in the USA:
* Do NOT apply to anywhere below the top 50 in the USA. I am pretty sure that any department below the top 50 has a very small department, turning all odds against you should you have any problems with your advisor/do not like that particular field anymore etc.
* Take all exams needed (GRE, TOEFL etc.) WAY BEFORE the deadlines. It's quite often the case that your scores are sent out to schools way after the promised dates. And you will simply be discarded because your application is not complete.
* As somebody already pointed out. Do not make the mistake I did. You will be far away from your home, and assuming you had a social network (not a virtual but a real one), you will lose all that. Go to a city that is LIVABLE. The US can get quite bad about this and many schools are located in very bad places to live in. No offense to anyone here, but this includes places like Princeton.
* Do not OVERAPPLY. I just made this verb up, and it means two things: (i) do not apply to a lot of schools, which will just be loss of money for you; (ii) do not apply to schools beyond your reach. This second part means that do not apply to UC Berkeley if you are a 3.50 (unless you have a very special reason such as a letter from a professor at UCB). I understand that in a lot of European countries it is the case that 3.50 is an amazing average, but since that is not the case here, the people just assume that you didn't do 'that' good.
Finally, I just wish good luck to everyone applying.
At 57:
ReplyDeleteNo, if you do theory, go to Chicago. If nothing else, you will enjoy being at the top of the department's political hierarchy.
(Conversely, don't consider Chicago unless you are a theorist.)
Question for all of you in the know...
ReplyDeleteUnfortunately switching to CS was a late decision for me (my undergraduate degree was in electrical engineering). I think this limited where I could get in even though I went to a highly ranked undergraduate school. Regardless, I ended up in a 30-40ish program. I am very happy here and enjoy it immensely. Sure, I would have more options in a higher program, but I have no regrets. Anyways, I am fairly certain I am going to be interested in an academic position in 4-5 more years and I'm curious if anyone has any recommendations on how to best set myself up for the best academic job possible. I realize I have next to no chance at getting a position at a school in the top 20, but is top 50 possible? What are the best steps I can take to maximize my chances for success. (By the way, hopefully this isn't too far off, but I am doing research in machine learning, not theory).