Final Project

8 minute read

Important note: This semester, the students can choose to submit either HW2 or the final project, whoever does both will get a 10pt bonus.

The goal of the project is to familiarize you with the state-of-the-art research being conducted in the field.

Milestones

  • 28th Feb: Choose a final project.
  • 31st Mar: Turn in the final report.

Choosing the project

  • Choose a paper from the list of suggested papers.
  • Papers that are not from the list are also allowed, as long as they are relevant to the course content. If you are interested in pursuing this option, please confirm with us.
  • If you would like to pursue a project that is not relevant to any specific paper, please confirm with us.
  • Comment yours and your partners’ name on the chosen paper in order to claim it.
  • A paper can only be claimed by a single team.

Tasks

Minimum viable project (MVP)

  • After reading the paper, try to identify the radius of works that are related to this paper. This will give you context about the contributions of this specific work.
  • For the papers that have ready implementations, try to reproduce the experiments and see if they remain consistent with the paper. Consult with us about papers with no implementation.
  • Write a ~10 page report. It should contain introduction to the problem, summary of related previous works, description, analysis & critical review of the paper, and demonstration of the results.
  • Find the limitations of the chosen work, and propose ways to alleviate them.
  • Copy-pasting from the paper will not get any points!

Research work

Try to do something beyond the MVP. For example, apply the method to a new problem, code the proposed improvements (even if it really doesn’t improve, the attempt is fine) and etc. It is not limited to this. Creativity and effort will be rewarded!

Evaluation

  • MVP: 60%
  • Research work: 40%

Submission

Create a zip file titled proj-id1_id2.zip (replace id1/id2 with your IDs).

The zip file should include:

  1. A single PDF document, report.pdf, containing your project report.
  2. A folder src/ containing all your own code.
  3. A README file (plain text/markdown) explaining:
    1. The structure of the code in the src/ folder: What is implemented in each package/module.
    2. Links to any external code you have used, including paper implementations.
    3. Steps to reproduce your results using this code: Where to get and place the data, how to run all the data processing steps, and etc.

Submission will be available here.