Slashdot Mirror
GRE CS Subject Test Prep?
coaxial asks: "So it's that time again to consider taking the GRE. While there are many resources on the web about the general test (mostly vocabulary building), the computer science subject test seems to be lacking. This is a shame, since this test covers pretty much everything in the undergrad curriculum. So I ask the grad student readers of Slashdot: what resources, besides the one book I've found, did they use."
Have you looked into prospective graduate schools yet? When I started looking at entrance requirements, almost all of them said that GRE CS test scores wouldn't even be considered in the application. I just focused on the "plain old" GRE, made sure to do well on the math section, and had no trouble getting into the grad school I wanted.
Because you can't spell "slaughter" without "laughter"
You will need it if you are relying on a single review book to get you through. I will guarantee you that reading one review book will not help much if you are not ready for it on your own.
The only thing a review book will accomplish is to trigger some of the memories of subjects that you learned. Even if you manage to read all of the material on the subject, it is unlikely that you will remember the exact question that they will ask you.
My advice is to review what topics are covered, and what they involve. Do not bother learning details, you can do it for one topic and still remember all of them. Be ready to go from your general understanding to the specific instance of the question. It is not that hard if you had a nice university CS education.
My studying: Reviewed Many-one versus turing reduction (I kept forgetting which one was which), took the practice test in the booklet the evening before the exam.
My score: 880 (I guesstimated all of the networking questions, as I have never seen networking before, did not bother studying it either. Looking back, studying would not have helped either.)
badness 10000
I took the exam in '99.
Forget the test prep packets, there is simply too much material that can be asked. Did you keep your old books? Get the books you had for any data structures or algorithms classes. The ones for the class after the introductory classes (ACM CS2 in educator speak) and then the last algorithms class you took. Those topics are the core of the exam. The other big sections were digital circuits and computer organization (memory, disk, processor, networking). You should have had a class on basic electronics and a class on operating systems that covered these topics. There will be a few questions about NP-completeness and formal languages.
Mathematical calculation is required. Stuff beyond 1+2 will be asked. You need to know linear algebra, matrices, combinatorics, graph theory, and basic calculus. The math is integrated with the CS topics.
Make sure you know the material before signing up. You won't be able to cram everything in a month or two. If you couldn't take the test today, with no preparation, and at least do decently, you aren't going to be able to make up for it by the fall.
What you can improve on is speed of recall. There are a LOT of questions, I think around 60 or 70. You must answer quickly and get it correct the first time. Good pacing is essential.
It's on a curve. You can miss several questions and still do well. I got the maximum score despite getting at least 4 or 5 questions wrong.
Basically, when you're a counter-terrorist, you've got to remember to *guard the bomb site*. I know one guy who was passed over because, when the bomb was planted, he was jerking off over in T spawn. He had no idea which site to go to.
Also, practice with the Deagle/AWP combo until you've really got that down. All the high level comps are dominated by that crap.
Just keep practicing until the test. I'm sure that those [GRE] guys will let you join.
Good luck!
# (/.);;
- : float -> float -> float =
I took the test this year, and I had a fairly good score (840/900, 91 percentile). A single test prep will *NOT* be able to cover all topics. Especially not to the depth that each needs to be covered. My suggestion- get a book on each topic. OS, graph theory (there is a *lot* of graph theory), complexity analysis, databases, compilers, etc. If you still have your undergrad books, you can use those. But do not expect to be able to study the test, like you can for the SAT or normal GRE. The number of topics, and the depth of knowledge needed in each, is just way too much.
This is not an easy test. I speak as someone who aced their SAT (1580), normal GRE (1510), and took almost 2 years worth of AP tests- this is the hardest test I've ever taken. If you aren't coming straight from undergrad (I wasn't), give yourself at least 3 months of intense studying. If you are, I'd still give yourself 2 months of targeted studying.
Best of luck to you. And on your school search as well.
I still have more fans than freaks. WTF is wrong with you people?
No amount of studying will prepare you if you don't know the subject. On the other hand, due to the constraints of the test (length and multiple choice) and the breadth of the field, it can't go particularly deeply into any topic.
The really important thing is to know what you have a good, current grasp of and what you need to learn/relearn/refresh: Do you remember what LALR stands for? Can you calculate the cost of a pipeline bubble or a branch misprediction? What information would you need to do so? How is depth-first/breadth-first related to stacks/queues related to LIFO/FIFO? What is the stack pointer? What is the frame pointer? What is an interrupt? What are the three primary elements of OOP? What is modus ponens/tolens (sp)? What kinds of race conditions are there? What is a critical section? What is the difference between a monitor and a semaphore? Define NP-complete. And so on.
I recommend picking up some practice tests (from the library if possible). Don't worry about taking them, per se, but go through them and make sure you remember something about the topics the questions cover. Think of it as sort of a checklist. If there's anything in the practice tests (or the list above) that you can't bring to mind, or feel fuzzy about, or never learned, go look it up (on the web, in your old textbooks, whatever) and brush up on it. Don't expect to learn about compilers if you've never learned them, but you should be able to answer questions about different categories of languages with regard to parsing requirements, for example. Go through the practice tests with a pen and paper and write yourself a list of topics for which you need greater clarity, then brush up on it from whatever materials you have handy. (You *did* save all your old textbooks, right?)
When I decided to go to grad school, it was because I wanted to be a professor and, in particular, a professor in a large, well-respected CS department (i.e. well-funded with a large pool of talented professors, grad students, and undergrads). Despite a B+ average as an undergrad (though my B.S. came from a well-respected CS department and I had taken many more courses than were required), I aimed high and applied to three highly-rated and well-respected universities for grad school. All of them required the CS GRE, though they simply used it as an easy first cut (anyone under a particular score, maybe 750 or 800, is out of the running immediately).
I was accepted by an Ivy League school with full funding (RA/TA, not fellowship). And I went. And I slogged my way through candidacy (which got me a Master's) with much friction and suffering. And I gave up and got a real job. And I'm far, far happier for it.
If you are convinced that you want to be a professor in CS, the only way to get there (in any CS department that expects any level of research from their faculty) is with a Ph.D. If you want to work at a serious research institution (e.g. Google or MS Research), get the Ph.D. If you just want the extra boost of a Master's for the money, get MS certifications instead (yes, I mean it, they actually mean something these days). If you feel that you can be better prepared as a computer scientist/software engineer/whatever by getting a Master's, you probably don't really know what you want and should take a closer look at your motivation. If you are treating grad school like the snooze button on the alarm clock of life, you are in good company and you should make sure that you are at least aiming for a Ph.D. even if you don't know if you'll finish it; the snooze button approach is all about leaving options open.
If it sounds like I don't think much of a Master's in CS, it's because I don't. I have one from an Ivy League school and while I learned a lot doing it and had fun and was exposed to all sorts of knowledge along the way, the four years I wound up spending on it have earned me a piece of paper that excuses me from not having any certifications yet. The best things that came out of my time there were social, not academic or career-related.
(On a side note, if you think you'd like to go to grad school because it's easier to meet a mate in school than in the real world, you're right. I doubt I'd be happily married now if not for grad school. See the snooze button approach above.)