I'm an undergrad studying math and CS at a US college. My plan is to go for a PhD in machine learning afterwards. Currently, I have two research experiences under my belt. The first experience has finished, and I have a strong LoR and a first-author publication at a small conference (along with two second-author workshops). The second experience is on-going.

My dilemma is that my second experience has very low probability of me publishing anything. In this experience, I work closely with a professor at my college, along with several lab mates, but she is an undergrad professor so her focus is on teaching. In other words, although she can write me a strong LoR (we know each other quite well), I basically have no hope for publishing (at least she hasn't brought this up by herself).

So, my question is, should I give up this opportunity to pursue one elsewhere that might result in a publication? Or should I stay with this opportunity and keep working hard and impressing this professor?

*Context: machine learning PhD application is quite fierce nowadays and more publications should help, but LoR also helps, so I'm confused.

1 Answer 1


In the US, good letters of recommendation are a must. Research experience, especially resulting in publication is definitely a plus, but since it is still a bit rare, you are probably already ahead of the curve there.

The "problem" with undergrad research is that, given the time (and maybe money) constraints it is hard to reach a conclusion on anything that is very significant. In research, you don't know the answers when you start and you are looking at the unknown. There is no schedule that you can apply to actually reach finality unless the problems are rather straightforward. The time between writing the paper and publishing it can also be long and indeterminate if the journal is reputable. Conference publication is a bit more time predictable, since the conference will be held on a certain date, but if you miss a deadline, the next opportunity might be a year away.

If you are doing something hard, then my recommendation would be to keep at it. People will, I think, reward you for that even if you don't reach a conclusion, provided that the advisor can praise your efforts and predict your future success.

I think you are already in a good position on this point, though doctoral admission is dependent on many things beyond the two you mention. But, in the US, letters are essential.

  • Of out curiosity, what are other things that you think are very important? Commented Apr 8, 2021 at 10:28
  • GPA, especially in major courses. Maybe the GRE for some places. What you say about yourself and how articulate you are.
    – Buffy
    Commented Apr 8, 2021 at 10:32
  • As you said, I might be "ahead of the curve", but I don't feel this way. In the following blog (timdettmers.com/2018/11/26/phd-applications) on machine learning PhD application (pay attention to the section "Can I get admitted to a top school?" and the Q & A at the bottom), it seems like it's extremely competitive to get into top schools. Maybe I'm confused between a "good" school and a "top" school. Anyway, there seems to be a lot of noise out there, which makes me very hard to evaluate my own progress. Commented Apr 8, 2021 at 11:20
  • 1
    You may be assigned an advisor at entry with little choice. But they are mostly just for course selection advice to help you get through the qualifying exams. But you should be allowed to pick your own dissertation advisor. Insist on it. But look around at several faculty before you do. Their fields. Their reputation generally. Their reputation for how they treat students. My biggest mistake in grad school was failing to change advisors since I was too introverted to speak up for myself. Talk to some other, more advanced, students as you go along. Pick a mentor or two.
    – Buffy
    Commented Apr 8, 2021 at 12:24
  • 1
    My main mentor was actually in a different subfield, but he was very helpful generally. My relationship with my eventual advisor was good but he was standoffish. You can normally wait a year or two (at least) before you need to ask to work with some individual.
    – Buffy
    Commented Apr 8, 2021 at 12:25

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .