Choosing shools right for my profile (PhD in Machine Learning)

[sinerged]

Hello all,

I'm applying for a PhD in Computer Science, specifically in Machine Learning, Artificial Intelligence,

and Robotics. My plan is to apply to two top-tier schools (e.g., MIT, Stanford, Berkeley), three or four

schools where I have a realistic chance of getting in given my credentials, and one or two safe

schools. However, I'm having a though time assessing which tier of schools is the "right" one for me.

At the moment I'm looking into Top 20-50 schools such as GeorgiaTech, Michigan (Ann Arbor),

Washington, North Carolina (Chapel Hill), UCSD, UCLA, among others. However, I'm not sure if those

ones are a little bit far-fetched for me. Anyone care to give a hand?

For those interested, here is a brief summary of my credentials. If you want to take a look at my SoP,

here is an old version (need to update it with info from a short internship I did this year).

Thanks for any help!

[newms]

Getting into any of the top 4 is chancy even for exceptional applicants, since there are so many applicants to these schools (I think MIT had around 3000 CS applicants last year - do the math, that's a really low admission rate if they take maybe 100 people?). I think you have a decent chance at top 20 schools and a good chance at top 50 schools, so I would focus at the other schools you have listed there. What's important is research fit, so try to look for programs that are doing research that matches well with your interests and your background. Good luck!

Edit: Having quickly looked over your SoP, something that stands out to me is how little you mention about your research interests. That should be an important part of your SoP. I would suggest that you focus more on your research interests (as well as related research you've already done) - such as what was it you did in surrogate modeling (especially if it relates nicely with what it is you want to research in grad school). What is it in machine learning that you want to research in grad school? There's a lot to machine learning, so being more specific will help. Perhaps you could talk about a specific machine learning problem that you would like to research. Also, you should talk a little more about why you would be a good fit at X university. What is it about the research there that interests you? Are there any professors there that you would want to work with and why? That is what will make your application stand out.

Edited November 15, 2011 by newms

[sinerged]

Thanks a lot for your answer, newms!

I'll definitively take into account your comments about my SoP. Reading some of the other threads in this forum

I realize my previous version was definitively lacking in the research interest department, and I've already started

to think how to improve that part. I saw some advice suggesting to actually begin the SoP stating my research

interests and how the fit with the program, and only then go on to explain why I'm qualified to do so. That's what

I'm planning to do with my SoP at the moment.

BTW, I do know how hard it is to get into Top 4. Last year I only applied to top-tier schools (MIT; CMU, Stanford,

Caltech, USC) and I didn't get accepted in any of those. That's why this year I want to make sure I pick more

realistic universities and I'm doing everything I can to improve my application package.

Thanks for your input, man

[OH YEAH]

I don't think you're applying to enough schools. Applying to 5 like you did last year is suicide, and even 8 schools isn't many in my opinion. I ended up applying for 12.

SoP suggestions:

- The vast majority of your SoP should be talking about your research experience.

- Make it clear in the first paragraph what you want to study. The adcomm is going to forward your application to the relevant professors, so don't make them wait until (literally) paragraph 6 to figure out where it should go.

- This is probably because you aren't a native speaker, but a lot of your SoP is "fluff". A sentence is "fluff" if it doesn't set you apart from the other people who apply or if it isn't backed up. For instance:

"An enormous enthusiasm for computer science and a large motivation to confront difficult long-term problems: these are the two qualities that best describe the approach I have taken towards my academic life."

Every student applying is enthusiastic about computer science. "Motivation to confront difficult long-term problems" describes your academic life? Can you back that up? What exactly does that *mean*, unqualified?

"My initial interest in computer science became a true passion after I enrolled as a M.Sc. student at the nationally renowned XXXX Institute, where I graduated with one of the highest grades in the history of the program. One experience from the program that I hold in high regard was attending Data Structures and Algorithms lectures under Professor XXXX. I was particularly captivated by the complexity of the subject and the difficulty of the assignments, and I used to stay up until late hours working out efficient and elegant solutions. This motivation rendered its fruits, and I was honored when the professor selected some of my solutions and test cases as benchmarks for future semesters. The experience taught me that in order to excel in computer programming, one not only needs a solid mathematical background, but also a large amount of creativity."

OK, you have a "true passion", but your evidence is that you took an intro level course? Millions of people take intro level courses in programming. The fact that it is difficult is why you are "truly passionate" about it? Why not study physics instead?

"As a research assistant I have always pursued projects that are challenging and that require the combination of various disciplines."

As opposed to non-challenging research projects? What does this sentence add? If your projects are challenging, it will be made clear by the descriptions you make of them.

"I am confident that my academic background, both as a student and as a research assistant, gives me a solid foundation to contribute to ongoing research projects at XXXXX. I am also aware of the high responsibility that an acceptance to this world-class university will entail and I would hold this honor with the highest commitment to hard work."

Mega cheesy. I wouldn't say anything like "the high responsibility", "world-class university", "hold this honor with the highest commitment to hard work" ... again, gotta back this stuff up. If your statement of purpose is good, there is no reason to promise things in your last sentence -- it'll be clear from your research that you take it very seriously.

[sinerged]

All right, that's definitively the kind of feedback I need in order to improve my SoP. Thanks, man! Would you be

willing to take a look at later versions of the SoP and give more feedback?

As for the number of applications, I'll try to take your suggestion into account, but I only ordered 7 official transcripts

and there is really not enough time to order more and get them before the deadlines. Some universities ask only

for online submission, so I guess that's my only shot.

[OH YEAH]

Well, in that case 8 is probably fine I'd be happy to read further drafts.

[sinerged]

Hey guys, could you give me your input about something?

The areas in current ML/AI research that I find the most interesting are the development of: 1) ML algorithms to

efficiently handle massive datasets, possibly using distributed computing techniques, and 2) efficient and effective

algorithms for planning and SLAM in robotics.

The problem with this is that I don't have direct research experience in those areas. My experience is relevant in the sense that

I have been working for years with neural networks, SVM, kriging, and other machine learning techniques for optimization

problems in engineering. However, I haven't worked in problems that would advance the state of the art in current ML;

instead, I've mostly been applying ML techniques. And in robotics (which happens to be the most appealing area for me)

I have even less direct experience.

Do you think this could be considered a weakness in my application, particularly in my SoP? How should I approach this aspect of

my application?

Edited November 18, 2011 by sinerged

[frenzydude]

I've mostly been applying ML techniques. And in robotics (which happens to be the most appealing area for me)

I have even less direct experience.

Do you think this could be considered a weakness in my application, particularly in my SoP? How should I approach this aspect of

my application?

Not really so. Different research groups have different strength areas. The lab where I work for instance, would easily accept people interested in applied machine learning, and not those interested advancing machine learning theory itself. So just focus on your strengths.

[OH YEAH]

And be sure not to say things like "Even though I don't have formal research experience, I have applied experience in X, Y, Z...." Just don't draw any attention to the fact.

[sinerged]

I just had a paper accepted for a conference and in my applications I listed its status as "In Review". Should I inform graduate departments of the recent change? If so, what is the best way to do so?

[CanuckBoy]

Yes you should. I had a journal paper that was accepted after I submitted my applications. I informed the schools and they asked me to update based on their procedures.

[DJLamar]

Just for reference and as a Georgia Tech undergrad alumnus, Georgia Tech is top 10, not just 20, and is certainly on a higher tier than Caltech and probably USC for computer science. Plus it's an even better school for robotics, computer vision, and machine learning (that general area is one of it's strong points for sure).