I am a first-year PhD student in machine learning at a reputable university in the UK. Over the past six months, my work has consisted of reading papers that I find interesting, studying textbooks, and experimenting with code. Although this may seem like a typical first-year experience, I have come to realize that my supervisor lacks proficiency in my field (I have had to explain fundamental concepts), which has made our meetings and my work less productive than anticipated.

Our meetings go something like this: I attempt to explain a paper I have read and then discuss possible incremental improvements, and the meeting would end with my supervisor telling me to go test my ideas. Although experimentation can be an essential part of research, coding every idea that I have is time-consuming. If my supervisor was more proficient in the field, I believe I would receive better direction and save a lot of time. They could immediately identify issues with my ideas or suggest other works that address the same problem.

While I understand that the PhD process should be student-led, I feel as though I don't have anyone to discuss ideas with.

I don't have experience with other PhD supervisors, so what is the role of a PhD supervisor? And what should I do about my situation?

  • 1
    Sounds pretty normal. "Coding every idea that I have is time-consuming" maybe, but that's what you're there for. You'll get faster (and more knowledgeable regarding which ideas are worth pursuing) over time.
    – cheersmate
    Mar 21, 2023 at 18:44
  • 10
    Can you explain what evidence you have that your supervisor lacks proficiency? You did not include any information that suggests that this is the case in your question.
    – d_b
    Mar 21, 2023 at 18:46
  • 2
    In any case, shouldn't you somehow arrange to switch advisors to one who is expert in the area of your desired projects? Mar 21, 2023 at 18:52
  • 1
    @user572780, based on some of your comments, is your project about applying machine learning in a particle physics context? Mar 21, 2023 at 20:29
  • 3
    That would really be a topic for a new question. (But in case you decide to post a new question, please check whether it has already been answered at Academia SE before.) Mar 21, 2023 at 21:32

3 Answers 3


If all the experts in a field know what the "result" will be, then there is absolutely no value in the research. Research is about finding the answers to questions we (we being all humanity) do not know the answers to.

If you read papers and have ideas about how to improve them, those ideas are only interesting and useful if they would require some sort of test. If your advisor recommends that you test your ideas, that's because that's how research works.

If you don't understand how to test your idea, that's something you should discuss with your supervisor. If you've tested your idea but aren't sure your test actually shows what you think it does, that's something you should discuss with your supervisor. If you have lots of different ideas to test, you might ask for advice on prioritizing them. Ultimately, though, you're going to have to test some ideas.

In other fields, there may be much higher barriers to testing a theory than writing some code: you may need to do weeks or months of experiments, or go through time-consuming regulatory steps for research in animals or with humans. In that case, you likely want to do a bit more refining before doing any single test.

  • 1
    @Bryan_Krause, I don't expect my supervisor to be to able to tell me which ideas are good. However, based on my experience, I believe someone with more experience would be able to immediately rule out some ideas.
    – user572780
    Mar 21, 2023 at 19:28
  • 1
    @user572780 Maybe, maybe not. Presumably you've not presented every possible idea, you've presented some ideas that seem reasonable to you. If you've read the work and "done your homework", you've already narrowed it down quite a bit. Your advisor may also see some learning value in doing the trial and error steps, even if they suspect some of your ideas are not likely to turn out. There's also the game theory aspect where the most interesting results might be when you try something unlikely to work and yet it works anyways.
    – Bryan Krause
    Mar 21, 2023 at 19:32
  • 5
    Maybe your advisor is incompetent, but your post hasn't convinced me of that, and I have some prior that they are not entirely incompetent in that they've been successfully hired to their current position.
    – Bryan Krause
    Mar 21, 2023 at 19:33
  • 1
    @Bryan_Krause, I don't think my post is meant to convince you of anything. Also, I am not claiming that my supervisor is incompetent. They might be the best particle physicist in the world. I am saying that they aren't proficient in my specific field as I am having to explain undergraduate-level concepts.
    – user572780
    Mar 21, 2023 at 19:44
  • 1
    @user572780 That's fine, but I'm going to write answers according to the information I have; I can't have information not in your post. Your mention of explaining undergraduate-level concepts didn't come until a comment written after I wrote my answer, though I'd wonder to what extent you're being asked to explain such concepts to demonstrate your understanding of them (perhaps in a specific context) versus educating your supervisor. If you're doing a PhD in machine learning, though, I can't say it makes any sense to me to choose a particle physicist as supervisor, how did that come about?
    – Bryan Krause
    Mar 21, 2023 at 20:01

I had the same experience when I did my Ph.D. in the UK. I worked on distributed computation and operating systems and neither my first nor second supervisor had a good understanding of what I was doing. To be frank, it was frustrating as there was no good academic exchange, debate, or collaborative work. A lot of my co-PhDs had the same feeling.

You will learn to live with it. You need to take the lead on your research earlier. You will decide what route you follow, and what ideas to try from the start (now). The supervisor's feedback will be mostly on the methodical and publishing side.

My supervisor's involvement improved when I had results. I guess, it will be similar on your end. As soon as you apply your machine learning research to the field of your supervisor, they'll get more involved.

My advice is to go to conferences and look for people who do something similar. Stay in contact with them and collaborate. This is something I should have done earlier in my PhD. Alternatively, you could look for an academic in your institute/uni that has more experience in machine learning / your area of interest and collaborate with them more.

Generally speaking: A supervisor should guide you a bit with your research, and help with how research is done. Ideally, you want one where the research group works in close alignment, but a lot of young supervisors get PhD students with very different focuses, which makes collaboration within the group very difficult. More seasoned supervisors tend to get PhD students that share a common focus. The second supervisor (in the UK) is often just a formality and is, in most cases, not involved in your research.


If you really want to be successful in academia, it is important that you have your own ideas and make progress on your own.

Your advisor seems to be open to letting you explore your own ideas and is not forcing you to do exactly what he wants, which is a fairly rare opportunity.

Also, unless your advisor is insecure about his lack of knowledge and prevents you from discussing ideas or learning from other groups, you can just try to collaborate or learn from others. Conferences, collaboration, whatever you like.

The impression that I get from reading your post is that your fear over your own ideas not working out makes you want to blame your advisor for not helping. If you're really so concerned that a specific idea is good, then go find a friend in the field and see what they think. (But show them in the form of a small written writeup of the idea so it's harder for them to steal it.)

  • My advisor doesn't prevent me from collaborating with others, it's the exact opposite, they encourage it. My biggest issue is that I don't have someone else to discuss ideas with. I tried to reach out to other researchers, but it seems that they were uninterested.
    – user572780
    Mar 28, 2023 at 11:58

You must log in to answer this question.

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