1

I have an undergrad thesis idea that involves transformers. I am trying to take the lead on the idea before I contact a professor about it - i.e. figuring out exactly how I'd make it happen, and if it is realistic. The problem is that because of covid, the largest research funding I can get from my university as an UG is $500. I read a paper about a type of transformer, and I would like to apply that to an area it hasn't been applied before (my interest area). However, it seems like transformer based research is expensive.

I am worried about two things:

  1. I don't want the professor to think I am idiot for suggesting an idea that might be big in scope / too expensive, but I truly do not know how to determine if it is too expensive / out of scope
  2. I do want to publish first author. If this idea is too expensive, am I allowed to ask if he has any other projects for me (I can also provide him with my other interests).
  • Somewhere in your department's course documentation, there should be a list of skills you're expected to demonstrate at each milestone stage of your thesis project. It might be under a name like "assessment criteria", "mark scheme", or "learning outcomes". You'll probably find that there are some of those skills that you can demonstrate by calculating the cost of your proposed research, and if it turns out to be too expensive, by revising your plans to make it cheaper while making a case that your results are still valid. – Daniel Hatton Sep 17 at 20:33
  • Since you have tagged computer science, are you referring to transformers, the deep learning architecture? If so, what kind of costs do you anticipate? If it's only computing power, working with a professor will probably give you access to their resources, and they will definitely be able to help you figure out if it is feasible and worth exploring. – GoodDeeds Sep 17 at 20:41
  • @GoodDeeds yes I am referring to the DL architecture, typically used in NLP. It is only computing power, but it might be a lot. Is this something I can explicitly ask him? Such as tell him my idea, and flat out ask him if it is realistic? I feel like gifted students should be able to determine this themselves – Emilio Sep 17 at 20:51
  • I don't think you need to have an exact idea, but knowing roughly where it lies in the spectrum between "can run on a desktop computer" versus "needs millions of dollars worth of computing power" can help. I believe (I may be wrong) the high computing cost mainly comes from the enormous amount of data used these days. Have you read papers related to your idea, and do you have an intuition as to how much data and so computing power (in terms of order of magnitude) you'd need? As long it's not ridiculously high, it should be enough to start a discussion. – GoodDeeds Sep 17 at 21:08
2

Is it my responsibility as an undergrad student to figure out if my thesis is affordable?

Usually, no. This is an example of where your faculty advisor should help you.

Do take a few minutes to research costs before discussing it with your advisor.

There might be exceptions if, based on your area of study, you were expected to know about costs already.

| improve this answer | |
1

It's great that you're taking the lead and refining your ideas to come up with a well-defined research proposal. Your prospective advisors will appreciate the effort and it will likely increase the chances of them accepting you as their student.

Cost is certainly an important aspect in determining the feasibility of a project. From your question, it appears that the main cost will be for computing power. It would be a good idea to attempt to estimate computational requirements to the best of your ability, by looking at relevant literature that you base your ideas on. If the costs are likely to be extremely high (as an extreme example, take the cost of training GPT-3), you could shelve the idea and think of something else.

However, no matter how carefully you go through this, it is always possible to miss something vital. Particularly, since you are undergraduate and it sounds like you are not yet very experienced in this field, you may miss factors to consider that are obvious to your professor. This is completely fine: you learn by working on projects, and it is the role of the supervisor to guide you through it. As long as you make a good faith effort and are enthusiastic to learn, it is very unlikely that they'll think you are an "idiot" because of this.

So, it would be a good idea to prepare well, but not wait too long to approach the professor. Their inputs can prevent you from wasting time on infeasible ideas, and also help improve your proposal. If they consider the idea feasible and interesting and accept your project, they would very likely cover most of the costs too by providing access to their lab's resources, so external funding, while welcome, may not be necessary.

Referring to your second question: yes, you can certainly discuss other ideas and ask for suggestions, or offer to join an existing project. It is unlikely to be unusual for professors to provide complete project ideas at the undergraduate level.

| improve this answer | |
-4

Maybe your best approach is to contact manufacturers who would have an interest in commercialisation and see what advice and resources you can get from them

| improve this answer | |
  • 4
    Comments coming in since your answer indicate that this does not refer to physical (electrical engineering) transformers, but to a computer science concept that uses code. Yeah, I went to a big hunk of iron and windings too... – Jon Custer Sep 17 at 21:00
  • My answer might still be valid, although now we are talking about software rather than hardware – Mikesplace Sep 18 at 4:15

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.