You are asked to carry out an original research project related to the course content. This description is intentionally very broad. Some possibilities include:

• Invent a new algorithm, architecture, or training objective based on the ideas in the course.
• Apply the ideas from this course to improve the evaluation of a machine learning system. (This might be a good option if you’re already using deep learning for your research, or are interested in doing so.)
• Throughout the course, we will sometimes speculate about why something happens, or what would happen in a given situation, and note that it hasn’t been rigorously tested. Rigorously testing these hypotheses would make a good project.
• Develop toy examples that clearly illustrate one of the ideas from the course or other ideas related to evaluation. (If you go this route, your setting should be significantly different from those that already exist.)

Collaboration

You should form teams of 2-4 students. The team does not have to be the same as for the paper presentations. Your final report should list the contributions of each team member.

Solo projects are allowed but not encouraged, and the final report may be graded by a TA rather than the instructor. The standards will be the same as for a group project.

Project Proposal (due 10 October)

Each team should write a short (1-2 page) research project proposal. It should include a description of a minimum viable project, some nice-to-haves if time allows, and a short review of related work. You don’t have to do what your project proposal says — the point of the proposal is mainly to have a plan and to make it easy for us to give you feedback. (But if you want to completely change direction, it’s a good idea to run it by the instructor first.)

Please submit your proposal through Moodle.

Final Report and Source Code (due 10 December)

At the end of class, you’ll hand in a project report, in the format of a machine learning workshop/conference paper in ICML format. We recommend the report be about 6-8 pages plus references, but we do not enforce any minimum or maximum length.

Please submit your project report and source code through Moodle.

Marking

80% of the marks will be given for meeting the requirements of the project and for the quality of the project proposal, presentation, and final report. This includes:

1. Abstract (5 points) that summarizes the main idea of the project and its contributions.
   1. Should be understandable to anyone in the course
   2. You don’t need to say everything you did, just what the main idea is and one or two takeaways.
2. Introduction (10 points) that states the problem being solved and why we might want to solve it.
3. Figure or diagram (10 points) that shows the overall idea. The idea is to make your paper more accessible, especially to readers who are starting by skimming your paper.
4. Formal description (15 points) of the model / algorithm / conjecture. Include at least one of: (a) an algorithm box, (b) equations describing your model, or (c) a theorem or formally stated conjecture. Highlight how your approach differs from existing work.