Project Proposal

Initial project proposal slides due at noon on 10/13. Your group will have a 15-minute meeting with the instructors during the regular classtime on 10/13 to discuss the proposal (see Piazza post for meeting sign-up spreadsheet). The final project proposal slides are then due at noon on 10/18.

The recommended (but not required) format for your project proposal slides is to have roughly one slide for each of the following:

• The problem you're trying to solve and why it's important
• Your proposed solution (you can have more than one proposed solution)
• Any existing related work
• How you intend to evaluate your solution (e.g., particular datasets, workloads, or systems); this may be the hardest for you to know ahead of time so do not worry if you don't have an evaluation plan in mind yet
• Timeline

During your 15-minute meeting with the instructors on 10/13, you will be given up to 8 minutes to present your slides, and the remaining time will be spent on giving feedback and answering your questions. We recommend that you use visuals as much as possible in your slides, instead of only text. Diagrams are especially helpful.

To help you come up with project ideas, we have compiled a list of potential projects that you can directly use for your final project. You can also look at this list of papers in the ML for Systems space for inspiration.

Midterm Report

The midterm report is due at noon on 11/3. Submit a written report (up to 3 pages) detailing the progress you've made so far. Please include the following:

• The problem you're trying to solve, why it's important, existing work, and a description of your solution. This should essentially be a written prose version of your proposal slides (assuming you haven't changed project topic).
• The progress you've made so far in terms of implementation and evaluation. Describe whether you've identified datasets or codebases that you will build on for evaluation, even if you don't yet have results to report. If do you do have evaluation results (e.g., plots), feel free to include them.
• Any challenges you are facing.
• A description of the remaining work and timeline for its completion.

In general, if you are facing challenges with the project, we encourage you to reach out to the instructors and TAs sooner rather than later.

Project Presentation

You will present your projects in a 12-minute timeframe on Wednesday 12/8. You will then receive feedback/respond to questions for 3 minutes.

There are a number of useful guides for presenting a research project (see this and this). On top of them, we propose you try to incorporate this flow in your presentation:

• What is the problem? Why is it interesting?
• What are prior solutions?
• What is your solution? Why can it be better than prior work?
• How do you evaluate these approaches? What do you find from the evaluation?
• What are the strengths and weaknesses to your approach? Can they be fixed in future work?
• Conclusion and main takeaways

Try to use as much visuals as you can and avoid slides with lots of text. Also make sure you can present your work within your time limit.

Final Report

You will also submit a conference style short paper (about 6 pages) on your project. You should use this LaTeX template, which you can copy into your own Overleaf project, or this Word template.

Generally, this flow is a good direction for writing:

1. Introduction: The introduction is basically a condensed version of the paper. Introduce the problem and motivation. Describe your solution at a high level and follow up with evaluation highlights. Finish with takeaways, strengths, weaknesses and future work.
2. Background: The aim of this section is to provide some background on your problem, if necessary. For example, if you are proposing a new congestion control algorithms, this section should be an introduction to congestion control.
3. Motivation: This section aims to give a concise high level understanding of the principle your solution stands on. It might not be necessary, depending on the problem/solution.
4. Design: In this section, you explain your solution in rigorous detail. e.g. If it is an algorithm, you provide a pseudo-code. Remember to highlight why your solution should (but might not in your evaluation) outperform existing work.
5. Evaluation: In this section you first explain how you intend to compare your solution to existing methods, and then present results. You should go over the results and explain them. Simply adding a plot or table, however powerful the results may be, does not give much insight to the reader
6. Related Work: This section should reference prior work related to your project. For someone who wants to understand your problem and its field, this section should include all necessary reading.
7. Conclusion: Here you go over the main takeaways and strengths/weaknesses of your method, and mention lines of future work.

The deadline for submitting the report is midnight, Wednesday 12/8. Remember to address any feedback you receive during the presentation to the extent that is possible.