CS 747: Course Project

You will undertake a research project as a part of this course. You have already taken a step towards the project by submitting a proposal. Your project is meant to be a short exercise in scientific research, aimed at furthering the understanding of some topic and/or engineering a new system. Since you only have about a month, keep the scope very limited. It is preferred that you do a good job executing a relatively small project, rather than a mediocre job in pursuit of a very ambitious goal.

It is a good strategy to "close the loop" as early as possible: that is, to integrate the different portions of your work early on. You can continue to iterate and refine thereafter. On the other hand, if you work on perfecting the individual portions of your project and leave it to the very end to get them to work together, it might be too late to take corrective action. If at any stage, you are not sure which portion of your work to prioritise, feel free to consult the instructor.

The primary deliverable from your project will be a report. In your submission, you must include supplementary material such as the code you have written and the data you have collected. We may examine such material as we verify your claims. However, the project will be evaluated solely based on the report. Do not postpone the writing to the very end: keep adding material as and when it is generated, and set aside the last few days before the deadline exclusively for the writing.

Make sure the report lists the names and roll numbers of all the members of your team. Also make sure that at least one team member uploads the report to Moodle by the submission deadline. If we receive multiple submissions from the same team, we will pick any one of the submissions for evaluation.

Your report must adhere to the principles of scientific writing. (1) Begin with an introduction that provides the context for your work, highlights the novelty of the investigation you have undertaken, and summarises the main findings and results. (2) Discuss related work from the literature; make sure your citations are all in standard format (see templates below). (3) Formally describe the problem you are solving, and then (4) describe the technical parts of your solution (for example, an algorithm you implemented). (5) Present your results, accompanied by a discussion of their significance. Good scientific papers provide enough details for the work to be reproduced. Hence, provide full details: parameter settings, libraries used for coding, etc. If there is too much clutter, you may consider moving some details to appendices. (6) Conclude the report with a high-level discussion of your project, identifying possibilities for future investigation. Check language and spelling. Acknowledge any help you may have received from colleagues.

Your project will be evaluated for novelty, significance, technical soundness, clarity, scholarship, and presentation. Although it is not mandatory, it is recommended that you adopt a template used for submitting a paper to conference such as AAAI, ICML, or NeurIPS. Go through a few papers from previous years' proceedings to get a sense of how to present your work. It will also be a good idea for you to stick to the prescribed page limit for conference papers (usually 6-8 pages).

Compile your report into a single pdf file called report.pdf, and place it in a directory called submission that also contains all the relevant code and data. Compress the directory into submission.tar.gz and submit it to Moodle by 11.55 p.m., Saturday, November 23. Your submission will not be evaluated (and will be given a score of zero) if it is not received by the deadline. Your team must be the same as in your proposal.