COMP 5118

• Project 45%
• Presentations 20%
• Paper Reviews 20%
• Class Participation 15%

Project (45%)

The project can be done individually or in groups (assessment will consider group size) and can be one of the following types:

1. New Research Idea: A prototype implementation of a new research idea that addresses a limitation in existing work or is a completely new idea inspired by your readings.
2. Experimental Study: A comprehensive experimental comparison and evaluation of existing work on a specific topic. The goal is to provide new insights not present in the original papers.
3. Survey: A paper that summarizes, categorizes, and provides new insights on a major research area. See this good example.
4. System Implementation & Reproducibility: Implement and reproduce the results of a system from a published paper. I have a number of systems in mind.

Project Deliverables:

• Project Proposal: Max 2 pages in ACM Format (LaTeX mandatory). Due: October 27 at 11:59 PM. Early submissions are STRONGLY encouraged. Submit via this form.
• Project Paper: Min 7 pages in ACM Format. Due: December 8 at 11:59 PM (hard deadline for late submissions: Dec 15). Submit via this form.
• Source Code: Publicly available on GitHub with a good README. Link must be in the project paper.

Presentations (20%)

Each presentation should be 30-45 minutes, followed by a 30-45 minute discussion. The presenter is responsible for leading the discussion.

Paper Reviews (20%)

Reviews are due at 11:00 AM on the day of the class via the Paper Review Submission Link. Format: Summary, 3+ strong points, 3+ weak points, and additional comments. Your two worst reviews will be dropped.

How to Write a Good Review:

1. Do not copy/paste from the paper. Use your own words.
2. Enumerate strong and weak points; do not write one large paragraph.
3. Justify your points. Why is a deep learning approach good *in this specific case*?
4. Elaborate! "Using templates" is not a strong point on its own. Explain why it is effective.
5. Focus on the technical substance, not just the writing style.
6. Avoid stating the obvious as a strong point (e.g., "they beat the state-of-the-art").

References: Reading a CS Paper, How to Read a Paper, and a real-world sample review.

Class Participation (15%)

This is a seminar-based class, meaning that your participation is essential. You are encouraged to ask questions, answer other students' questions, and give comments on the papers we discuss.