Preparing for GRE CS

[addict]

Hello everyone,

 

How long does it take to prepare for a good AGRE score (90% +)? I have around 75% in undergrad and 3.85/4 (expected) in MS.

 

From where should I start preparing?

[CalmLogic]

Will you be applying for a PhD? Is your MS that you are finishing in CS? Did you already take undergraduate CS courses in data structures, algorithms, computer architecture, etc.? If the answer to all these questions is yes, then I would focus more on research. You will have already established by your GPA in undergraduate and graduate CS courses that you have the requisite CS background to do research. I would try to use your MS thesis as a way to get a paper (even a "poster paper" of 2 pages) accepted at an international conference or journal. (Assuming you are applying for a PhD in CS, research ability is ultimately what an admissions committee is looking for, so why not give them exactly what they want by doing peer-reviewed research?)

 

Having said that, it certainly doesn't hurt to the take GRE CS exam, especially since the score doesn't have to be reported along with the general GRE scores. At a minimum, studying for it will help you when it comes time to taking PhD qualifying exams. For UT Austin and Rutgers, taking the GRE CS exam seems almost mandatory, at least for those who are completing a BS rather than an MS. On the other hand, the time you spend studying for the GRECS exam may be better spent on research, so studying for the exam as much as possible may be a bad idea if it interferes with getting published.

 

(If you live in India, my understanding is that the next time you can take it is in November.)

 

As far as how to start studying for it, I would start by looking at the current ETS booklet, taking the exam in that booklet, and seeing which areas you need to brush up the most. Keep in mind that the exams in the ETS booklets only cover a subset of the GRECS syllabus, which is mentioned in the current booklet.

 

Of course, there is much more advice in the old posts. A more recent post with some advice on how to get started:

 

http://www.www.urch.com/forums/gre-computer-science/81548-subject-gre-cs.html

 

Since you apparently already have the advantage of having taken graduate courses on this subject matter, you will be more likely to score above 90 percent but that is far from assured even with a lot of practice. Unlike the general GREs, it is far more difficult to predict how one will do on the GRECS. Most of those who have posted in this forum were more likely to be shocked with dismay than to be shocked with happy surprise after getting their GRECS score. Unlike a PhD qualifying exam, you can't choose which areas you want to focus on.

[addict]

Thank you CalmLogic for your reply.

 

Yes, I will apply for PhD. My undergrad background is in CS and currently I am doing my MS in Computer Engineering in a moderate rank US university ( rank varies within 50-60 in Computer Engineering).

 

Well, I agree with you that research and publication are more important for PhD admission especially for a candidate who would apply after MS. But the universities where I want to apply seems very choosy and most of them especially top rankers prefer subject GRE. The list is:

 

Rank within 10

Stanford University

UC Berkeley

Caltech

CMU

Michigan Ann Arbor

 

Rank between 11-20

U Southern California

UC San Diego

UC Los Angeles

Maryland and College Park

U Washington Seattle

 

Rank between 21-30

Not decided

 

Rank between 30-40

UC Santa Barbara

UC Davis

UC Irvine

U Colorado Boulder

 

The list has a bias towards California as I want to shift there. And here arises the problem. I think UC system is more choosy than other similar rank universities (correct me if I am wrong). Stanford, Berkeley looks impossible without GRE CS (even if I have publication and good GPA). Even USC, UCSD, UCLA somewhat fall in this category I think (again please correct me if I am wrong).

 

Please reply.

[CalmLogic]

But the universities where I want to apply seems very choosy and most of them especially top rankers prefer subject GRE.

What evidence do you have of that? Many CS department websites will say the GRECS exam is recommended, but that's a very, very general statement that doesn't necessarily reflect the feelings of the current admissions committee, at least when one compares the import of publications to GRECS scores.

 

For example, regarding CMU:

 

At CMU we receive hundreds of applications each year from 4.0 GPA students who have never done research. These are all put into the high risk pile and are subsequently rejected.

 

Your score on the GRE will be largely ignored by the top schools, particularly if you are coming from a top school. At CMU we look at the Verbal GRE score only when the applicant is not a native speaker of English. We look at the Subject test score only when we have an application from a school whose CS department we’re not familiar with.

 

http://www.cs.cmu.edu/~harchol/gradschooltalk.pdf

I assume the professors at CMU are familiar with the top 100 CS departments in the United States. BTW, a funny Dilbert cartoon about CMU:

 

CMU and Dilbert

 

Regarding Berkeley:

What’s the secret to being admitted to UC Berkeley? A: Our Ph.D. admissions are based primarily on research potential. If you can demonstrate in your application that you have strong research potential, it will make a difference. The best way to demonstrate that is to do a lot of interesting research. If you've written any published papers, include them in your application. Videos or web pointers to software systems -- especially if they demonstrate novel research --will also help a lot. The point is to demonstrate that you love research; such people are especially welcome. Also, strong letters from people who know your work well help. Of course, if the person writing the letter is someone our faculty knows, that letter will be given extra weight, but it is better to get a strong letter from someone who knows your work well than a lukewarm letter from someone famous who doesn’t know your work well.

 

I'd like to emphasize that this advice is not just for students applying to Berkeley, but for students applying to any research university CS program. I know first hand that evidence of strong research ability is welcomed by faculty and Berkeley and CMU, and based on discussions with faculty at a wide range of universities, I can assure you that the same is true at other top computer science universities throughout the US, Canada, and Western Europe.

 

Tygar:* FAQ for Applicants

I think the best thing you can do for yourself is:

 

1. Publish brilliance.

 

2. Repeat step 1 :)

 

Also, see what MDK and I have said here:

http://www.www.urch.com/forums/computer-science-admissions/82622-please-evaluate-my-profile-suggest-i-am-loss.html#post534465

 

 

Stanford, Berkeley looks impossible without GRE CS (even if I have publication and good GPA). Even USC, UCSD, UCLA somewhat fall in this category I think (again please correct me if I am wrong).

 

I would suggest contacting graduate admissions advisors at some of the schools you are interested in, asking them about the import of publications vs. GRECS scores for PhD admission. You have nothing to lose by asking. For example, one could ask, how many of your accepted PhD applicants reported a GRECS score? I don't know if they would answer that, but even if you got them to speculate, that would be very helpful.

[MDK]

Recently, I came by a statistics for the GRE CS test takers.

During the period from Jul 2003 to Jun 2006, 7200 students have taken the GRE CS. Their average score was 717 and the standard deviation was 92.

Recall that 717 is translated to around 45%. So, this small number of test takers didn't get great scores. In fact from a quick reading for the numbers I can tell that any score above 780 is really great. Still it's not good enough for the top schools that you talk about ( UCs, Stanford...etc).

My point is: if we assumed that 2400 students have taken the test every year (7200/3) then the top 10% of them would be 240 students. As I have read from Berkeley website that they have accepted 100 students most of them achieving above 90% in the GRE CS. that left us with 140 students, since University of Washington also states that all its students achieved > 90% and I assume they accepted not less than 50 students. that would leave us with 90 students. So there is no way in hell that all the other prestigious universities shared the remaining 90 students.

What I am trying to say is: I believe that the universities are bluffing the things out in order to make free advertisement for themselves. I firmly believe that the number of all of the examinees that took > 90% can't sufficiently fill the available positions in top schools.

Thus, most of the universities have to compensate the GRE CS with other factors such Papers, GPA, LORs.

 

In my opinion, you should work on your research instead of preparing for the GRE CS because if you do a little math you will find that the chances of getting a high score ( upper 90s) are minor.

And remember that: A bird in the hand is worth two in the bush

[CalmLogic]

Interesting findings, MDK. It is pretty unbelievable that the University of Washington is reporting the average GRECS score to be 91 percent while their average quantitative GRE score is only 90 percent: Admission Requirements We are left to believe that those who scored below 80 percent would not dare report their scores to the University of Washington.

 

The only CS department statistics I really trust regarding GRECS are those by Rutgers. I have mentioned their statistics before, but I don't know if they apply to accepted MS applicants or include accepted PhD applicants:

 

The mean GRE scores for a recent entering class of students were: Verbal 580, Quantitative 782, and Computer Science Subject Test 752; the mean TOEFL score was 268 (minimum 213 required by the university).

 

Computer Science, Rutgers University: M.S. Program in Computer Science

My guess is that the statistics above include PhD applicants. Someone who was accepted into their PhD program with full funding had published research at an IEEE conference. She got only a 740 on the quantitative GRE and a 750 on the GRECS exam:

 

I talked with my adviser about re-taking the GRE and he said that is the least important factor in the application and I shouldn't waste time on it. He said as long as it is enough to cross the preliminary filtering, no-one even cares about it.

I was concerned about my subject GRE, which I had written in 2003 and hadn't done well, I had scored 750. Apparently it didn't make a difference.

 

http://www.www.urch.com/forums/admissions-results/46872-phd-cs-rutgers.html

(However, the person with the above profile is a woman, which can be a significant advantage since the female/male ratio at most CS departments is obviously very low.)

 

I realize Rutgers isn't at the top like UT Austin, but even UT Austin seems to be overrated regarding the GRECS:

 

for fall-2005, 34 of the selected 62 odd candidates [at UT Austin] didn't have AGRE scores.

 

http://www.www.urch.com/forums/gre-computer-science/12301-material-computer-science-subject-test.html

That's pretty significant when you consider that it is often assumed that most people accepted to UT Austin have taken the GRECS and reported their scores. If the above finding is still accurate, that means that 45 percent of the people who are accepted to UT Austin have reported GRECS scores. This percentage would probably be much lower at most other top schools since UT Austin recommends the GRECS exam more than most like USC.

 

I would also mention the painfully obvious: The students who are applying for a PhD at a top school are less likely to report scores below 60 or 70 percent just as MS applicants are less likely to report scores below 50 percent. So some professors on the admissions committee may not fully understand how hard it is to get a high score on the GRECS.

 

BTW, some different scenarios:

 

1) Student studies for the GRECS exam. Gets 80th percentile.

 

2) Student works on research with or without co-authors. Gets paper accepted at a good conference.

 

3) Student works on research but doesn't get publication accepted. However, student writes paper and:

 

- has professor review paper and praise it highly in a recommendation letter

- attaches paper to application

 

 

Obviously, scenario 2 is generally best for a PhD application. Scenario 1 and 3 may be debatable as to which is best.

[MDK]

The only CS department statistics I really trust regarding GRECS are those by Rutgers. I have mentioned their statistics before, but I don't know if they apply to accepted MS applicants or include accepted PhD applicants:

I also believe the University of Texas at Austin since their average two years ago was similar to Rutgers:

http://www.www.urch.com/forums/computer-science-admissions/80363-agre-770-please-evaluate-my-list-2.html#post524888

 

In fact I am more convinced by the idea of preparing a paper for publishing rather than preparing for the GRE CS. That's why I didn't schedule to retake it again after getting 770. In fact I studied for it only a month (combined with general GRE ) and I remember that I went to the test without reviewing all of the ( Titanium Bits ) imagine that.

After taking the test, I knew that if I studied for the test (2 or 3 months as most of the people here do) I will be able to get > 85%. But I am not planning to do that, simply because I decided to give the effort on my senior project so that I can publish a paper after finishing it , rather than preparing again for a test that it's not safe. In addition, when you join a graduate school you need a research capabilities more than general knowledge of all computer science spectrum.

On the other hand, the only thing that I may benefit from my GRE CS is that no one in the states knows how the computer science department in my country like, so it may ( and stressing on may ) help

[addict]

Thank you calmLogic and MDK for your reply.