I am a computer science and math double major, and I am applying to computer science PhD programs this fall. I have a 3.9 GPA and have done research with 3 professors during my time in undergrad. I have one publication and three poster presentations, and I will hopefully have at least one more publication before I apply (but due to problems in the research group, that might not happen). I can get two great letters of recommendation, but the third might be lackluster. I go to a regionally known liberal arts school, but it is most certainly not a household name.

With my qualifications out of the way, I want to study machine learning. My prior research was not in machine learning per se, but it did involve aspects of it. I have done work in bioinformatics, computational statistics, and computer vision. My professors are not especially well known in their fields, but they do publish frequently. What tier programs should I apply to? My advisor thinks I should apply roughly on the threshold of top/mid tier programs, but I would like more opinions.

On a tangent: Some schools have both computer science and machine learning PhD programs. Should I apply to both? And do some schools admit generally with no specialty? If so, would I have a better chance with them since machine learning is so competitive?


At least in the top-4 (I'm an ML PhD in one of them and helped pre-screening candidates) Machine Learning has become incredibly competitive. We almost always asked for 1 publication (if possible in ML, although not always). Third recommendation letter is probably irrelevant, the two most relevant things were the most informative letter (the professor that knows you best) and indication that you can become an independent creative researcher (1st author papers, out-of-ordinary projects or accomplishments). GPA almost didn't matter. Note that I've heard some of these criteria are different outside top-tier.

Applying to CS will likely increase your chances of getting in. I know many students doing ML in non-ML labs, but this also has downsides and you have to find an advisor that accepts you. Not applying to Reinforcement Learning within Machine Learning will also increase your chances (at least last year).

It's hard to know if you have good chances of getting into top schools, but if you have money and time I would apply widely, it doesn't seem impossible. I applied to 10 schools and have friends that applied to 15. It's pretty stochastic, I got into 3 of the top 4 but got rejected by multiple others.

Focus on doing good research in ML, if possible a workshop at a top conference, and impressing your current advisor.

Best of luck!

  • Thank you for a thoughtful answer! I am going to apply to as many programs as finances allow in hopes of mitigating the mentioned stochasticity. One final question: as you said, I do have a better chance getting into CS, so for the schools with CS and ML programs, should I apply to both if they allow? Or is that just a waste of energy and money? – Overfit Jul 26 '20 at 18:18
  • one thing to keep in mind is that in PhDs you apply to professors more than schools. Yes, the admission is done at department level, but you should mention professors you're interested in. Thus, check whether each school has professors that excite you. You should also definetely look for exciting professors in less famous universities. – etal Jul 27 '20 at 19:45

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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