1 of 6

Timeline and key dates

  • Nov. 1: submit title/abstract proposal (online form)
  • Nov. 8: (by/optionally) receive project feedback
  • Nov. 13: (optionally) revise title/abstract
  • Dec. 7: poster and revised abstract submission
  • Dec. 11: poster presentations (2 sessions, 1.5h each)
  • Dec. 15: final report submission

9.520/6.860: Project Guidelines

2 of 6

Project types

Project should be in direct contact with the course!

Theory (extension of a derivation, study of an algorithm)

​

Comparisons/Critical evaluations (of group of methods)

Benchmarks ...

​

Application (of SL/ML principles to your own area/task)

Focus on the learning problem, regularization, ...

​

Review (research areas, summarize/distil larger/technical papers)�

Implementation: code something fun

*If the project is not focused on implementation, any programming language can be used.

*You can use any dataset you like (use several to show generality if data is not the focus)

​

​

9.520/6.860: Project Guidelines

3 of 6

General guidelines

  • Individual projects
    • collaborations not encouraged.
    • exceptions of teams of 2 should be discussed in advance.

​

  • Notify in advance (proposal/abstract) if this is:
    • a project for another course
    • an extension of another project
    • part of your research
    • part of your published work

​

​

Plagiarism will be taken seriously.

​

Not following these rules implies not getting the project evaluated.

9.520/6.860: Project Guidelines

4 of 6

Proposal

  • When: by Nov. 1

​

​

  • What: title + abstract (15 lines/300 words), 1-2 refs (if needed)

​

  • Will not be graded, should be informative

​

  • You might (!) receive feedback for (for focus or change of topic etc) by Nov. 08. If not, you are good to go!

​

  • Revised (optionally, after feedback) proposal: Nov. 13

9.520/6.860: Project Guidelines

5 of 6

Evaluation criteria

  1. Achievements: are the goals achieved? If not, how much work has been done?
  2. Methods: were they chosen appropriately?
  3. Context and connection to existing work
  4. Technical correctness: e.g., experiments OK, result interpretation, theoretical correctness, ...

​

​

  • Poster presentation clarity/quality

​

9.520/6.860: Project Guidelines

6 of 6

Project development

During development: interact with the TAs

  • Office hours
  • E-mail (for logistics)
  • After class

​

Report (NO REPORT)

  • NIPS format
    • Style file and template: http://www.mit.edu/~9.520/fall17/#projects
    • 7 pages maximum (including figures, references, appendices)

​

Deliverables (Dec. 7):

  • Poster
  • Revised abstract using single-page NIPS abstract template.

​

​

​

​

9.520/6.860: Project Guidelines