12

I graduated from my MSc two and a half years ago, so I have been outside of the University environment for a while.

I have recently started working on some novel machine learning techniques. Writing an article by myself and submitting it to a prestigious journal is no joke. I'm wondering: how likely is it to get supported from a University, or by specific researchers that are working in a University (for example PhD or postdoc students) with my work? I envision them becoming co-authors of the paper. Would they benefit from this? For example can they get some credits for correcting the paper and being a co-author even if I'm not a PhD student?

Clarifications:

  • I'm developing my own technique, independent from my employer. It's a general topic (regularization) so it's not directly applied to an industrial task.
  • I want to apply for a doctorate, but my GPA and MSc thesis grade weren't brilliant. Hence, the need to publish an article before applying for the PhD - but also for the pleasure of writing about innovative research, of course
2
  • 2
    Please don’t write answers in comments. It bypasses our quality measures by not having voting (both up and down) available on comments, as well as other problems detailed on meta. Comments are for clarifying and improving the question; please don’t use them for other purposes. Existing answers in comments and the ensuing discussions have been moved to chat.
    – cag51
    Mar 1 at 15:38
  • Submit your paper to a peer-reviewed conference near your city Mar 2 at 5:57
34

I agree with Wetenchaap's answer and just want to elaborate the picture from our (university) side. Occasionally people write me and suggest something along the lines you envision, so, in general, yes, it is possible. However, my personal experience with this type of collaboration is mostly negative so far, and thus now I'd think twice before venturing into such projects again.

So, what's the problem? Usually it boils down to expectations and work management. People outside academia often misjudge what is considered "interesting" or "publishable", and what kind of work is typically required to compose a paper. More often than not getting a full picture cools down their motivation. Next, since people typically have jobs and do research in their spare time, months of efforts can easily end up with a short "sorry, piled up at work" message. Thus, investing into someone else's idea requires trust that lies on some firm foundation.

Dealing with students and colleagues is much easier. Students need it for their graduation works and (possibly) for their future careers if they consider staying in academia. Colleagues do research with me as a part of their normal daily activities. Outside people are unpredictable: today they believe that publishing a joint paper is good for them, but tomorrow they can easily change their mind.

1
  • 10
    It's good to point out that the root cause is in "research in spare time" more than "research outside academia".
    – MSalters
    Mar 1 at 10:02
13

You don't need "support" from a university to publish. What you need is work good enough to satisfy journal reviewers and editors. You can always publish as an "Independent Researcher" without other affiliation.

However, if you want help in the research or writing, then you need to make some personal contact with an interested party. Being in Amsterdam gives you some options as there are universities within reach and someone there might be interested if you pay them a visit.

Blind emails to people aren't likely to be very productive.

And, given that you are interested in a doctorate, you might try to apply to a few places and include your current work as "work in progress" in your CV. Of course you will need to meet the admission requirements, and I'd think that some letters of recommendation in your case might be helpful if you can arrange them.

Post-docs probably have their own projects and little time. And working with an "independent researcher" isn't unheard of for an academic, but you need to find a way to make contact. Even if there is someone who is trusted and can introduce you to someone else.

7

For university researchers, it would definitely be beneficial to become co-authors of a paper published at a top venue. This is their bread and butter. They are unlikely to get explicit credits in any form, but a publication is its own reward, even as a co-author.

It may be nontrivial to convince university researchers to work with you, though. Maybe their funding compels them to work exclusively on specific projects. Even if they're completely free to work on whatever topic they want, you will still need to convince them that your ideas will be worthy of publication, before they will be willing to invest significant amounts of time on it. But if your ideas are good enough, being convincing is possible.

Reach out. You may get "no" for an answer, and you may get that answer often. But you may also get a "yes". Only one way to find out.

2
  • 2
    "They are unlikely to get explicit credits in any form, but a publication is its own reward, even as a co-author." I am confused by this sentence, authorship is an explicit credit, no? Mar 1 at 4:36
  • Fair enough. I interpreted the phrase "can they get some credits" in the original post as referring to course credits, like here: en.wikipedia.org/wiki/Course_credit Mar 1 at 8:28
3

The other answers are missing a key point. For you, writing a paper is stupid. It's a big waste of time that will take forever assuming you can find someone to help you, which you probably can't. It will give you academic currency that's basically useless for you because you are not an academic yet.

Let's talk about alternatives. You have a new/modified method you think has improved performance on a problem in stats/ML (regularization). A perfectly valid way to show your work is by writing it up informally (in a blog post or on twitter) and developing a software package (most likely in R or as a sklearn module in python).

For example, let's look at UMAP, a relatively recent addition to the nonlinear dimensional reduction toolkit:

  • It started on twitter as an idea in 2017, with an associated software package.
  • It showed up on the conference circuit in 2018 (I first heard about it as a scipy talk)
  • It turned into an arxiv paper in 2018, which currently has something like 2500 citations. It's a great piece of software and lots of people use it every day.

The advantage here is that the process moves quickly, it requires skills that you already have (not academic writing, which you probably don't because academic writing is a very specific skill that nobody should bother having if you don't need it), and it's also a tangible thing that you can put on your PhD applications. The disadvantage is that you might get "scooped", which I won't bother to define because it's idiotic in 2021 to think that for stats/ML the software package and associated writeup is itself not the key research output.

3

I publish on the topic in question.

If you were to write a paper, the first step would be to get a good understanding of the relevant literature. In doing so, you'll end up having a good idea of who publishes work similar to the method you are proposing. I would recommend looking those people up and contacting them: an email stating that you're working on something similar and would be interested in chatting. Set up a short video call, run your idea past them, and see if they're interested.

Key would be to know your audience: you'll likely want a 2nd/3rd year PhD student. Younger and they might not have any idea what they're doing. Any older, and they're either busy writing up, or a full time busy grownup postdoc/prof.

Even if no one is willing to work with you, you're likely to get useful feedback on your idea that may improve it.

PS, I strongly disagree with @Libor's statement as to writing a paper being a `stupid' idea. At the research group I am at in ML, a publication at a top conference is a prerequisite for applying for a PhD position. That's a fairly common requirement in the field. A blog post/non-peer-reviewed arxiv article/code would have to be extremely impactful to be accepted in place of that (i.e. a higher threshold).

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.