jjsakurai

Depends on your knowledge - for instance if you're comfortable with probability, you probably don't need to take Concepts of Prob. I'd take Stochastic Processes, the two linear modeling classes and sampling surveys. If I wasn't comfortable with basic prob/stats, then I'd take concepts in prob/stats and the two linear modeling classes. Of course, if you can, you should try to take all the classes you've listed. As to math classes - definitely make sure you're very very good with linear algebra. Also if you've space, real analysis and measure theoretic probability. They won't be directly useful to pratical ML research but are helpful in giving you a conceptual framework and if you want to do some theory research in ML.

Probability in general unless you're really interested in crypto in which case number theory.

@mathgeek282: That could just be an effect of better students going to higher ranked places. But I agree with your general point that while not as important as faculty fit, rankings are important. In addition to the prestige value, a higher ranked place also on average has better students which is extremely important if you want to have good collaborations. Plus if your interests change or your choice for an advisor leaves, a higher ranked place is more likely to have a good person you can work with.

Yes they were American. And while rankings are a shallow indicator, for someone not familiar with a particular subfield, it's one of the few indicators they can and do rely on. Of course what you do in your graduate study is vastly more important but unless you're a superstar in the field, going to a better ranked program can help you get an interview for a faculty position. And by rankings I mean, the ranking of the program in the particular subfield. So if you're doing NLP at USC/JHU/UMD, you'll be better off than doing it at say MIT. Similarly for architecture if Wisconsin has a better rep in that area compared to MIT.

They're roughly similar in terms of ranking. Unless there are specific research interests that I can pursue only at Michigan, I'd personally go for Columbia. It's in NYC which will be advantageous in terms of job opportunities plus the Ivy tag could help a bit.

Yes - you can do that.

This is not universally true. Check out Section 3.1 of the Grad School Talk(http://www.cs.cmu.edu/~harchol/gradschooltalk.pdf). Probably depends a lot on the person reading your app but my guess is that for most people it will make a difference (although the bias may be subconscious).

If you do apply again, I'd recommend expanding your list to include schools that are not just the top few.

GPA does matter. While no one will care if your GPA is say 3.8+, if it's less than say 3.6, people do start to notice. It depends a lot on the person(s) reading your app, but 3.05 is certainly very low for a top PhD program and can sink your app even if you've other mitigating factors (like excellent recs and research) unless you've a good excuse for the 3.05 (like illness) or if your bad grades are mostly in areas unrelated to your field of study. Also while too much attention to coursework can be detrimental to research, GPA does correlate well with how well you understood the material in a course. And understanding course material in the area you're working in is very important for doing research in that area. That said, I've seen people with GPAs as low as 3.3 get accepted to top programs based on the quality of their recs and research.

Next year might be too soon. Given that you applied to 8 universities and got in 1, I'd say that it's unlikely your app fell through the cracks. You want to send a better app in the next time around. If you can get research experience and collect more recs preferably from well known people in your area, then that will improve your app. I remember a person who posted on the GradCafe results who had applied and got rejected at most places in 2011 and got accepted to most (including most of the Big 4) in 2012 (I think they were in Gaming or something related). So it's definitely possible. As for self-funding in PhDs, it definitely makes a difference if it's some sort of fellowship. I'm not sure how much though.

Nah...Pittsburgh is not exactly a top-notch city and is in the middle of nowhere. Plus the startup culture in SV can be infectious. But if you're not that into startups, then you won't miss SV. And while Pittsburgh is no SF or NY, it's not that bad to spend 1-2 years in.

While I understand that an MS is something that can be used to differentiate from the pack and all things being equal, a recruiter will pick an MS over just a BSc. But all things are not equal. The background between different candidates is usually very different. In terms of interviews, you'll rarely find two candidates who're exactly the same. Also places like Google also filter based on the GPA and what university one went too. In addition, especially for tech startups, even having a degree is not as important as it used to be compared to your past experience (unless the startup is doing some serious engineering), much less an MS.

@Pauli: Fair enough although my & my friends' experiences differ greatly. After getting the initial interview (the chances of which showed no bias towards MS students), the only thing that seemed to matter was your background in the area you were looking to work in and how well you did your interview.

@Pauli: Could you give examples? AFAIK (and this corresponds with the experience of my friends), there's no such push. You don't for instance require an MS to find tech job at the big tech firms nor at startups. This is true both in theory and in practice. For my class, the only difference b/w MS and BSc was that the starting salary of people with MS was $5-10K higher.

Btw, Zoop - another option you could try is working at University affiliated labs. MIT and CMU Robotics for instance, employ several people who are not students at the university and IMHO, that would be a better option than working in the industry or even in an industrial research lab (unless of course, you're getting a chance to work with someone great at the industrial research lab).

You're misunderstanding me. I didn't say industrial experience couldn't be useful in research. I'm saying that AFAIK, the adcoms don't care too much for industrial experience compared to research experience at a good university. So for instance, an applicant who did say 2 years of good research at a top 20 university would be looked at more favorably than a candidate who spend say 10 years at a company like Google. I've several friends working in ML/IR at Google and Microsoft and most of their work involves implementation, tuning various algorithm parameters, optimizing the algorithm to run in a distributed setting, etc. Research in ML/IR at good universities does have those components but what is valued there is the conceptual work that you do. (Of course, if you're working in a good industrial research lab, that helps your case a lot as I mentioned before but industrial research labs are a very small part of the tech industry in general) So yes, industrial experience is useful but I've not seen a single student at a top 10 university who got there on the basis of industrial experience alone. Invariably, they had research experience at a university. The essay portion is one of the least important parts of the app. It can hurt you if you do it badly. I rarely helps you.

It's really hard to game things like that. First of all - Princeton has many more theory people as a percentage of their department so your prior odds of a non-theory prof recruiting you are not that high even though fewer non-theory students will apply to Princeton compared to the MIT/CMU/Stanford/UCB schools. Secondly, there will need to be a reason for a non-theory prof. to hire you - which means you will need to have done significant research in that area and have recs from people the adcom knows about. MS admissions are usually much easier but you usually have to pay too. Honestly - you still have 3 years left! Get excellent grades in the next 3 years and your freshman grades will make little difference. Also do tons of research in the areas you like. Try to get a few papers published. You might try to find summer research opportunities with well known advisors. Even if you're from a school that's not that well known, you'll still have a decent shot at the top universities. Btw, only the grades in your major (and specifically the subarea you're interested in matter). So your physics grades won't matter. Similarly if you're a theory applicant, then your Systems grades will not matter much either. A 3.8 is plenty enough.

CMU definitely. This is a no-brainer. The only thing Stanford has going for it is that the Silicon Valley is next door and the weather is nice. Research-wise, CMU is a lot bigger and better. Reputation-wise, CMU is as big a name in Tech as Stanford. It's true that the average person on the street will be more likely to have heard of Stanford but honestly - that's really irrelevant and you don't want to base your decision on that. Any recruiter will know how good CMU is. Plus if you want to go for a PhD later, there are a lot of funded MS Robotics students who stay on for a PhD. Even if you get a TAship at Stanford, RAship is so much better than a TAship. Plus all the tech recruiters come to CMU. What Stanford can offer is many more opportunities to be a part of a startup while you're pursuing your MS. There are many more Stanford students (and not just in CS) I've seen doing a startup compared to CMU and MIT - even while they're still at school.

Engineering and some science fields - yes. CS (except for Systems and PL) is not one of those AFAIK. With only an undergrad degree, it's very very hard to get the sort of position where you'll work on the kind of research you'll do in grad school (again Systems and PL are the exception). Also, an industrial letter is worth much less than an academic letter - see point 4 here (http://matt-welsh.bl...rad-school.html) written by Matt Welsh when he was a CS prof at Harvard. He also says this in the comments - This not to say that industrial experience is not valuable. Just that in terms of CS grad school admissions, unless you're interested in Systems/PL - and I guess HCI too, the industrial experience will not be that important (unless you really created something noteworthy which people in your field have heard about) and is generally not a substitute for doing research at a university.

I disagree with the advice to do a couple of years in the industry. Unless you want to go into a field like systems or PL, industrial experience will be of no help. If you however go to a research lab and manage to do research with a well-known member of the community, that helps your case a lot. As to the original questions - it may be possible for someone of your background to get an admit at a top 20 PhD school. I'd definitely recommend that you apply if you're sure your recs will be fabulous. Some schools also offer a funded MS while at many schools, once you get admitted, you can find profs willing to fund you after a semester or two. I'd also recommend that you check out Canadian schools - typically they're cheaper and they've some amazing schools like Toronto up there.

Probably because applicants to higher ranked schools are a self-selecting bunch - so the average applicant to a place like MIT is much better than the one to a place like Duke. In addition, a smaller size would drive down the acceptance rate.

When I was applying for a PhD, I did email potential supervisors to see if they were hiring any students (as that would have factored in as to whether I applied to that school). Almost all of the profs I contacted replied back. I think if you express a desire to work with them and ask something along the lines of whether your background would be ok for the work you want to do with them, it shouldn't viewed on negatively. Keep any emails you send short and to the point. Don't ask for them to chance your admission or anything like that - they get dozens of such emails everyday. Also bear in mind that the odds of a professor responding are still not very high as they are usually swamped with emails. You might also want to ask this question at http://academia.stackexchange.com/ . They might have someone with knowledge of this.

Second the suggestion for applying to an HCI program. One thing I'd recommend if you do decide to take the HCI route is to have a good portfolio.

Do you want a PhD or an MS? Also what do you exactly want to do with the degree - do you want to head into research, be a software engineer or just learn more about how to program? For instance, if your goal is to be a much better programmer, then you don't really need a degree for that.

For the PhD, I think the admissions have gone out already. You might be admitted of the waitlist though - if they have one.