Overview

The Course Project is worth a significant portion of your grade. It offers you the chance to apply your newly acquired skills towards an in-depth application. There are two options for the course project: choosing one of the sample / default final projects, and choosing your own topic. We strongly recommend you do the course project in a team. Teams can be up to 5 people. Keep in mind that larger teams will be expected to do correspondingly larger projects. You should only form a 3+ person team if you are planning to do an ambitious project where every team member will have a large contribution.

Important dates

  • Course project proposal: due April 17.
  • Course project milestone: due May 15.
  • The final presentations / demo day will be held in the end of June (TBA)
  • Final submission: due TBA

The project proposal and milestone are small, and are only worth a small amount of credit for submitting them on time. However, you cannot use late days for them. You can use up to three late days for the course project. However, note that if you are in a team, pushing the deadline back a day takes one late day per person (see the grading page).

Default projects, datasets and interesting papers

To inspire project ideas we have listed some interesting datasets, papers as well as well as provided some sample default projects that you can choose from.

You can see more information here.

Project requirements

The project proposal is only for groups doing the choose-your-own final project. Please submit one proposal url per team (any member of your team can submit) to the Google Form by 11:59pm on April 17th. You can not use late days for the project proposal. Your project proposal can be short (a single page). It should have the following headings:

Team: List the members of your team (names and Faculty Numbers) Mentor: If someone has already agreed to be your mentor, list them. Otherwise, we will assign a mentor to your team.

Problem Description (1-2 sentences): What is the problem that you will be investigating? Data (1-3 sentences): What data will you use? If you are collecting new datasets, how do you plan to collect it?

Methodology/Algorithm (2-4 sentences): What method or algorithm are you proposing? If there are existing implementations, will you use them and how? How do you plan to improve or modify such implementations?

Related Work (3+ prior works): Which papers will you read to inform your understanding of the problem, and the appropriate methodology to tackle it?

Evaluation Plan (2-3 sentences): How will you evaluate your results? Qualitatively, what kind of results do you expect (e.g. plots or figures)? Quantitatively, what kind of analysis will you use to evaluate and/or compare your results (e.g. what performance metrics or statistical tests)?

Minimal Requirements: Does your project meet the following requirements? • Dataset with at least 10,000 labeled examples (more will be needed for some tasks like machine translation and 100-way classification). • Dataset can be completely collected by the project milestone due date. • Task is feasible: either prior work on the dataset exists or a human can get good accuracy on it. • You have identified an automatic (i.e., can be computed by a computer) evaluation metric for the task.

If not, justify why your project will still be feasible.

Practical Tips

You can see some practical tips on how to manage your deep learning experiments – for example, using Git to mange your code, TMUX to manage your VM sessions, and how to monitor your CPU and GPU usage.