weninger

Most have been admitted... but many have not.

By... not terribly difficult, I mean, if you meet the stated requirement (GPA, TOFEL, SPEAK, GRE or whatever) and you seem like a serious candidate, you are likely to be admitted (same with Stanford, Berkeley, etc. MCS from what I understand). PhD and MS with thesis is identical admissions process, that is, very very selective (source: I review applicants (now that this cat is out of the bag... I shall not be answering specific questions.) )

The most important thing that you should understand is that at UIUC, MS in CS != MCS. MCS is a professional masters program. Coursework only, no thesis, no research. MS in CS at UIUC (and probably Purdue) is admitted and treated exactly like the PhD (with obvious MS != PhD differences). Many PhD students get an MS on the way. So, this is incomparable... sorry :-)

MS and PhD have the same admissions process and requirements.

The first few rounds of admissions are out. The next rounds will come as folks start figuring out their funding sources and graduation rates, etc. I don't think I'm allowed to say more than that ;-)

Its not terribly difficult to get into the MCS program at UIUC. As a current PhD candidate at UIUC, my understanding is that if you have reasonable grades, GRE and SPEAK score, then you stand a reasonable chance at admission. There is never funding given for MCS students, and they don't receive an adviser, so its the department is more comfortable to open the doors :-) Same at Stanford and Berkeley from what I understand. However, this is not to be confused with the MS in C, i.e., MS != MCS. MS has the same admissions standards and process at PhD.

To follow up on my last post, more to the point. You've got a good list there. I would certainly add CMU and UDub. Depending on what you are interested there are really great researchers a WUSTL, UPenn, and UNC-CH. For PhD students my advice is: find a list of professors first, then consider the school. For MS/MCS students: always pick a top 5 school if you can.

The Microsoft Academic Search is great for some things. But this is not one of them. H-index and paper numbers are not the best indicators of research quality. Most salient point -- People can write lots of papers in crappy conferences. A much better indicator is funding amounts and sources. That is, NSF funds are competitive and very rigorously reviewed by top scientists. If a university (in America) has 3 awards for ML research vs 1 or none, then its obvious where the quality research is.

Its tough to say because at UIUC MS/PhDs are evaluated together, and as you can expect, we are highly selective. We try take to take the top few students from each country. For example, in the last four years my research area (5 profs) have admitted one student from IIT bombay who had a paper in a top conference. I can't say for sure without seeing your full application... but I will say that its tough even for the top students at IIT etc.

I am a PhD student at UIUC, and while you're correct that admission is competitive for PhD/MS, its not so bad for MCS. MCS is also not funded at UIUC, probably not funded anywhere. If you like research topics why not consider MS/PhD?

You'll have no problem getting into a decent grad program. Maybe Stanford, maybe not. But I would certainly apply for top 5 with your credentials. If you like Jure's stuff, we'd be interested to have you at UIUC. Why don't you email me and we can talk about it. Google can find me. :-)

Folks, As you guys already know, you are an awesome community. I used theGradCafe as a resource when I was applying and now I suppose I should make a donation: I won the NSFGRFP and NDSEG fellowships (Computer Science). My essays are available online for your use. My essays address many of the opinions that you guys seem to have been debating (references, broader impacts, etc). http://www.cs.illino...1/research.html (at the bottom) Also, I'm on the admissions committee at my school (Top 5 in CS). Let me know if you have any admissions questions.

The goal here, reiterating my point from earlier, is that research publications are meant for an audience. It is often the case that researchers will write papers to venues they believe they can be accepted, and thats fine; I think that its best if you submit your papers where interested people will read them. These two often go hand-in-hand -- but not always. If you read the related work, which I assume you did otherwise you'd have no business writing a research paper, it follows that people of similar interests will read the same venues. Therefore, succinctly, you should submit to the venue where the majority of the related work appears. Good luck!

I am a PhD candidate in Data Sciences at a Top 5 CS school. Most of the researchers here write their code in Matlab or C++/Java. For our purposes there is no need for software engineering, language semantics, arch design, etc. Math Theory and stats is important, but the Bachelors CS stuff isn't necessary so long as you have a cursory understanding of programming. While I didn't read everyones post before mine, I would tend to say that you can apply to Data Sciences PhD without a CS degree. My office mate has her BS in Math, and she does fine. Good luck to you!

I have posted my winning NDSEG and NSF proposals on my Web site. Enjoy: http://www.cs.illinois.edu/homes/weninge1/research.html

You are correct. The letters of recommendation are always due a bit later than the actual application. No need to worry. Make sure you get your stuff in on time, and make sure your advisors meet their deadline too!

Your NSF fellowship application is self contained and bears no weight on your actual research plan. For example, if you propose to study nano-particles and end up studying fashion merchandising, NSF will not care. The field of study is to route your proposal to experts in the field. For example, a proposal on nano-particles should at least be reviewed by particle physics, etc. That said, a sociologist may not care about your STEM education plan, but an educator might not be able to appreciate your research experience. If I were you, I would propose your STEM education research (which you seem excited about) and list STEM education as your field of study. Of course, you'll have to write your research experience and personal statement so it is accessible and can therefore be appreciated by your reviewer. Ultimately, if you win, you can still study fashion merchandising and NSF won't care. Good luck!

I agree, I missed NIPS for AI. -- These were off the top of my head, which is why I didn't list any Systems or Theory conferences As for weights: conferences are typically more highly regarded than workshops and symposiums. But this is not always the case; for example, the WebDB Workshop is a very highly regarded workshop and I would send a paper there before many other database conferences. It all depends on your specific field. The general rule is that you want to have your work read by as many people as possible. Ergo, try to find the venue that is most read in your area. Good luck!

This is an easy answer: do research. Graduate school is all about cultivating young researchers. If you can prove your research aptitude then you have a far better chance at getting into your "dream" school. I, for one, would take a student who is first author on a publication in a highly reputed conference than a 4.0 GPA with 1600 GRE. There is a problem with this though... the risk/reward ratio for research is much higher than a job. For example, if you are hired for a job at a company for, say, the summer, then that job goes on your resume and you get credit. On the other hand, if you spend a summer performing some research and it gets rejected (like the vast majority) then you've nothing to put on the resume besides "paper submitted". That said, I think I would still rather take the student who has "paper submitted" than the student who had an internship, because, like I said, research aptitude is the most important part of graduate school, and, therefore, anything you can do to demonstrate that aptitude will help your admissions probability. Good luck! Tim Weninger

This entirely depends on what topic you are publishing in. Computer Science is a very broad, interdisciplinary field, so there are literally hundreds of publication venues. One thing that is different from most other fields (eg. physics, chemistry, biology) is that the most widely read publications in computer science are actually conference proceedings. Journal papers are more for archival purposes and completed work. That said, you need to pick out a conference in the subtopic that you wish to submit to. For example, I do web data mining, so I typically submit to: WWW, CIKM, WSDM, SIGKDD -- these conferences are pretty well spread throughout the year so no matter when I finish my work I can submit within a reasonable time frame. If you are in Databases: SIGMOD, VLDB, ICDE Data Mining: SIGKDD, PKDD, SDM, PAKDD, ICDM Information Retrieval: SIGIR, CIKM Artificial Intelligence: AAAI/IJCAI, UAI, GECCO Machine Learning: ICML, AAAI/IJCAI Natural Language Processing: ACL, NAACL, EMNLP Social Networks: SIGKDD, ICWSM, WWW and so on... Look at the venues that your most cited references are from and consider publishing in those venues. If you read the papers from those venues then it is likely that the interested people will read your paper. The hard part isn't picking the conference, the hard part is being accepted. A reputable conference will not have an acceptance rate higher than 30% (ICDM has 6-7% acceptance rate, and I hear SIGGraph is similar) Good luck!

I think that you have a good shot at getting into a PhD program. I am interested in the actual journal and conference publications that you have. By just saying "international conference" that can mean anything. What the professors will look at is your authorship role (first or second author) and what journal/conference it is in. If you're a good fit, and that paper is in a good conference, then you'll be fine. Your GRE score is good enough, but US departments don't really care. If you reply with your conferences and authorship roles then I can give you a much better estimate... maybe recruit you myself Good luck!

Three things: First, most top schools do not distinguish their applicants between MSc and PhD. They just accept the top X number of candidates as graduates students. Once in, most PhD students, in my estimation, actually do get a MSc degree on the way to a PhD. They do this because PhD is (as you say) a long road, and if they don't finish they still want something to show for their efforts. So if I were you I would apply for a PhD program (which will likely get you funded) and then get the MSc on the way. If you find that you don't want to spend the extra 2-3 years then don't and you've lost nothing. Second, quals vary from department to department. Typically, in America, you are required to take 35 or so course-credit hours. The qual does not waive that requirement. So, although PhD candidacy is based on passing the qual, you cannot graduate PhD without passing the dozen or so courses. Third, of your list: 1) Berkeley 2) Stanford ...are fantastic ML schools. You might add CMU and UIUC and Cornell to that list too. 3) Columbia 4) Princeton ... are not ranked as highly, but if you find a professor that invests in your particular line of ML research then you'll be fine I do not have any opinion of European Schools, except that I've heard a lot from MaxPlank in Germany and Edinburgh in Scotland -- although I am not specifically aware of their ML prowess. Best of luck!

Of course you can just go down the list of top schools and keep picking. For what its worth, and I know this is a long shot, but Kansas State University continually gets first (or second) place in the machine vision competition in AAAI. I'm not sure why this is... but you may add it to your list of schools. Otherwise, the top 5 (Stanford, Berkeley, MIT, CMU, UIUC) are all top notch in the area. I am currently at UIUC and Roth is great for ML and Forsyth is great for CV. I'd ask your current professors for their opinion too. Good luck!

Dan, First of all, I'm sure that you realize the importance of making progress in the pure and applied mathematics fields. Otherwise you probably wouldn't be interested. The field of combinatorics has direct implications in computer science which has very real implications in the real world. So, if you can link your proposed research with an open optimization problem like gene sequencing or graph search/traversal, then you should have no problem convincing the reviewers of the impacts of your research. Secondly, the reviewers of your proposal are very likely in the same field as you. They know the implications of your work, they are just looking for you to say it. Thirdly, my winning essays in computer science may (or may not) be of some help. They're not pure math... but they're close They are available on from my research page at: http://www.cs.illinois.edu/homes/weninge1/ Tim Weninger

I have to disagree from most of what the other posters here say. The GRE subject test is not used during admissions. The regular GRE is not used in admissions for top 5 universities. Source: I've talked to several admissions folks at my university (UIUC) and others.