I'm faced with a rather tough dilemma :-

I had completed M.Tech in CS 2 years back, completing my dissertation in the field of VLSI Testing. While I liked my work, I don't wish to go back to pursue my PhD in that - I've been strongly wanting to pursue a theoretical course (in approximation/online algorithms) as means for obtaining my doctoral degree.

However, since I've had no research experience in TCS(Theoretical Computer Science), I'm afraid that would hurt my chances of getting admission to a reasonably good school in the US - whereas, the research experience in VLSI (as well as the LORs from my adviser/committee members, who are well known in the field of VLSI) would have helped me get into a good program (but my enthusiasm in that field has already petered out).

That's why, I wanted to hear from people who have successfully migrated from their initial research field (in UG/MS level) and have been able to embark on a completely different field for their PhD - how did you explain the reason for your shift in the SOP, whether that affected their chances in getting into a top school etc. Also, for any academics browsing the question, what would be your take on it - do you always prefer students who have a background matching your interests only ?

  • 5
    $\begingroup$ this question feels somewhat on the border of being off topic, but I'm not sure. $\endgroup$ Sep 24, 2011 at 5:51
  • $\begingroup$ Its difficult to accept any of the three answers as the "best" answer, as they all answer my query from different angles - hence I've upvoted all of them, and keeping this question open (in case someone else also wants to shed light on a different perspective) $\endgroup$
    – TCSGrad
    Sep 26, 2011 at 19:36

3 Answers 3


In some departments, PhD students are only admitted if a faculty member is willing and ready to advise/fund them. You may have a harder time being admitted to these departments to be a theory PhD student if you have not already done some theory research; there's not much incentive for individual faculty members to take risks on students who are less experienced in their fields, especially when they have to spend hard-earned grant money to do it.

Other departments expect their PhD students to float for a year or so, taking a broad swath of classes and maybe working as a teaching assistant, before settling on a particular research area and advisor. These departments may be more willing to admit smart students who are switching areas, or just don't know what area they want to work in, because the risk is amortized across the entire department.

Even for the latter departments, though, you want to cast yourself in the best light. Your statement of purpose should describe your past research experiences in detail (demonstrating the intellectual maturity, independence, stubbornness, etc. required to do research anywhere), before explaining why your current research interests are leaning more toward theory. In other words, what Dana said.

A few points of personal anecdotal evidence:

  • Over the last ten years, my department has moved away from the second admission model toward the first. Our students get into research faster, but we admit fewer generally smart students who don't already have focused research interests/experience.

  • David Eppstein may correct me, but the second admission model is how I got into theoretical computer science research. I applied to UC Irvine to study software engineering. (I had been a professional software engineer for a few years, so given my crappy undergrad GPA, that was the only thing I could credibly claim.) By the end of the first year, during which I worked as a TA and did not have a formal advisor, I realized that theory was a much better fit.

  • Amir Nayyeri, one of my most successful PhD students, started his PhD program at UIUC working on sensor networks, because that's what he did as an undergraduate. But once he was in the door, he decided that he'd rather do theory, took and aced some graduate theory classes, and started doing unpaid theory research while also working as an RA for his sensor-network advisor. So switching fields is possible, even in departments where students are pigeonholed from day one.

  • $\begingroup$ Thanks a lot for your expository post !! One query though - I've heard (and read) that switching advisers once admitted tends to make the student rather unpopular with the faculty (as his former adviser is left with an empty seat that could not be filled before next year) - is that somewhat true ? $\endgroup$
    – TCSGrad
    Sep 25, 2011 at 17:40
  • 4
    $\begingroup$ There is some danger, but only if (1) the student tries to keep the move secret and/or (2) the former advisor is a jerk. If you've made an agreement to work as someone's RA for a semester or a year, honor that agreement. If you're thinking of switching advisors, keep everything out in the open with both advisors. On the other hand, you're at grad school to get your degree, not to serve some faculty member's research agenda. In any case, it's best to discuss these things with potential advisors before agreeing to work with (never "for") them. $\endgroup$
    – Jeffε
    Sep 26, 2011 at 8:03

I recently served on the TCS grad admissions committee of MIT, and I can give you my angle on things: When I'm reading applications, what I'm looking for is excellent people who strongly want to pursue research in TCS. The combination of intrinsic ability and strong desire is, in my opinion, the best way to predict success. After all, it is our job, once someone becomes a grad student, to help them transform their potential into actual achievements, through courses, advising, etc.

Here are some important points :

  • Before grad school is a very early stage to change fields. Many people shift then. It's acceptable and expected.
  • In your application, explain your changing fields (so the readers understand why the rec letter writers are not from TCS, why you took lots of non-theory courses, etc). Talk about how you learned TCS, what appeals to you in TCS, demonstrate that you know the basics of the field.
  • If you have strong letters from your supervisors, it still counts for you, even if they are from a different field (strong = from a professor who knows you well and has great things to say about you).
  • If you have research experience (especially if there's some theoretical angle to it), it still counts for you, even if it's in a different field.
  • If you have good grades, it still counts for you, even if they are for courses from a different field. For TCS, math classes are important.

My impression is that admissions committees are generally looking for "strong" students where "strong" is largely defined via prior research experience. i.e. they especially want to know if you've had the experience of doing research, were successful at it and have a good idea if this is something you really want. The letters are important as experienced researchers' evaluation of your research ability and potential and grades give some indication of overall academic aptitude.

So I would claim that the specific area you're interested in is not AS important as having these indications of being a strong researcher through prior research, recommendations and grades (especially in relevant courses). After all it is entirely expected that you might do something different than what you first intended and change directions during grad school.

But obviously the research area is taken into account, particularly when it comes to matching students with faculty (a department would not want to admit 10 theory students if there is only one theory faculty currently taking students).

Also I would think it helps if you can show even some small project in the intended area, even if it's an unpublished ArXiv paper, just to show seriousness about the new area.

I personally know someone who was admitted as a networking student and within one year switched and ended up as a very successful theory researcher.

So in short I don't think the change of area will significantly hurt the chances of admission as long as you are an otherwise strong candidate.

(NOTE: This is mostly based on what I saw of the CMU admission process)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.