2

Last semester in my master's program, I came up with an algorithm that I thought was an interesting improvement on another algorithm we learned in class. I brought it to my professor, but he never got back to me with any kind of feedback. I'd like to start looking at it again (just as a personal project) to see if it is a legitimate improvement upon current methods, but I'm not sure where to start. Is there any specific progress I should make when looking into this project before I bring it to a professor to get their quick thoughts? For example, should I run a few tests and present them? Maybe try to write an entire paper in the style of a piece of academic research?

To be clear, I don't need a professor to invest tons of time into me (I know they're very busy), I just don't know the general structure of how to go about conducting a research project and would want to make something I bring them worth their time.

Also, would it be acceptable to contact any professor in the school even if I'm not in a class of theirs? Currently the only professor of a class on algorithms I'm taking is also my academic advisor, but he doesn't meet students one on one.

Thank you!

3
  • 3
    Just getting stuck on "he doesn't meet students one on one." Is that usual for masters students? That would be a problem for me as a student.
    – Boba Fit
    Commented Feb 15, 2023 at 16:45
  • 1
    There is some context here: Looks like this is a $100K taught master's where the professors send you to the TAs, the TAs are non-responsive, and "even for career events, tickets will be gone by the time I open the email to RSVP." Yikes.
    – cag51
    Commented Feb 15, 2023 at 16:59
  • I completely forgot I posted that before, but yes that is the context. It's gotten slightly better this semester but I think my expectations were a bit too high overall.
    – DomM
    Commented Feb 15, 2023 at 18:06

1 Answer 1

2

To make progress on this, you will certainly need to test whether your algorithm is better than the state-of-the-art. So I would recommend that you think about an experiment -- ideally, leading to one "money plot" -- that would show unequivocally whether your algorithm is better (and by how much). This probably involves taking a widely-used implementation of a standard algorithm and comparing it with your own implementation of your algorithm. This will let you convince yourself and others that you're on to something here. And you'll probably learn a lot by doing this.

Ideally, you could discuss this test with a professor (or even a PhD student in the appropriate subfield) before spending time on it. It may be, for example, that the "textbook algorithm" you learned in class is only of historical or conceptual importance, and so even a successful experiment would not really be interesting. Or the dataset you are thinking of using is not a good choice. Or your algorithm, while marginally more accurate, will be prohibitively slow. So getting a sanity check before spending time on this could save you from wasting time.

But either way, if you make the plot you have in mind and it still seems promising, then it's certainly appropriate to ask a professor for a brief meeting. Check our archives for advice on requesting this meeting -- a very concise, fact-dense e-mail is your friend here.

I would not recommend writing up a full paper at this stage. If professors are not keen to meet with you, they will be even less keen to read an unsolicited, unpublished manuscript. Good luck.

2
  • 1
    By the way, this is how things "should work" -- and usually do work -- but it sounds like your institution is not the best. That doesn't really change my answer, but I realize that you may have more trouble getting meetings than is typical.
    – cag51
    Commented Feb 15, 2023 at 17:01
  • 2
    Even if there is no research value (as described in the second paragraph), taking the steps suggested in the first paragraph are likely a useful exercise to learn the "process", as long as the time investment is reasonable (perhaps tens of hours for a student, rather than hundreds).
    – Bryan Krause
    Commented Feb 15, 2023 at 17:17

You must log in to answer this question.

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