Like the assignments, the final Research Project for this class should be done in groups of 2-3.
For the class research project, you will be required to have 3 meetings as a group with the instructor (Gilbert) and then give a final presentation to the class. These meetings are intended to help keep you on track!
The first meeting you will have with Gilbert is a project pitch meeting. Your group should come up with 2-3 ideas. Be creative! You have some amazing classmates. You will be able to come up with amazing ideas by mashing together unexpected combinations of backgrounds. Brainstorm!
The purpose of this meeting is to get feedback on which ideas seem most promising. The meeting will be more valuable if you have multiple ideas that you genuinely like. The meeting will be less valuable if you've already 100% decided what your project is going to be.
What is your thesis?
Why should people care?
What is the problem? (concisely)
What is your new idea? (concisely)
What are your contributions/claims?
Do these contributions/claims demonstrate your thesis?
Are these contributions/claims proven by your methods: proof, implementation, experiment, evaluation, etc.?
Are your contributions/claims objective? (i.e. using your methods, will a colleague who does not like you or your idea nonetheless be forced to concede that your contributions/claims are justified?)
What are your methods?
What experiments/evaluations are you running?
Why do these methods suffice to demonstrate your contributions/claims?
Is there a simpler method to demonstrate the same claim?
Your research project should involve differentiation and/or probability in some way.
Your research project should involve some interesting systems aspect. Merely using PyTorch or Tensorflow to do something is not sufficient. What is your interesting technique?
Your research project should highlight some application. Don't get so far into system building that you lose sight of the reason for building systems.
You will be required to write up a 2-page project proposal for review and discussion with the instructor (Gilbert). The purpose of this proposal is to demonstrate that you have put together a good plan for doing your research project. A good plan must have the following:
You need to have a draft thesis, contributions, and methods (see above). These can change depending on how your research progresses. But you need to have clearly thought about these aspects of a research project.
You need to have made a basic effort to look up and describe related work. Gilbert will try to help you improve this, but you need to start looking on your own! Has someone already done your project? (That can be ok for a class project, but you should know whether it's the case)
You need to have a plan. Specifically, I want your proposals to include a timeline. The project will probably not go according to plan. However, trying to draft a timeline will (1) be a sanity check on whether you can accomplish your project by the end of the quarter and (2) force you to not procrastinate.
If you think "oh gosh, there's no way I could do this in a quarter" then revisit your methods (see above). Is there a simpler and more economical method for demonstrating your claims?
At this point, you will take your proposal and edit it into a draft final paper (probably at least 4 pages at this point). You will have a meeting with the instructor (Gilbert). At this meeting you will discuss any obstacles you've encountered, and get suggestions on how to make the most efficient use of your last 2 weeks. The basic purpose of this meeting is to help prevent complete disasters.
At the end of the quarter, you will have a "research paper" (6-8 pages) to turn in. You will make a brief oral presentation on this work (probably around 10 minutes) in front of the class.
Your final project will be assessed according to the above criteria. You will get the points if you made a good effort at doing research. You will not lose points because your research project didn't happen to go well. Research is uncertain. That's OK!
Project Pitch Meetings (Jan. 10&11) - Signup Sheet
Proposal Meetings (week of Jan 29) - (signup todo)
Preliminary Results Meetings (week of Feb 19) - (signup todo)