Projects

Practical work is an essential part of learning. We offer you an opportunity to test your newly acquired machine learning skills against real-life problems. Please, read this page carefully and only then ask questions (in Slack, #project).

Key dates:

• Team formation (team size is 2 - 4 students), until October 14 (Friday)
• Intermediate presentations November 14 - 16 (5 points out of 25 points)
• Final presentations December 12 - 14 (20 points out of 25 points)

Choosing a project

The first step is to find a project that is aligned with your interests. Below are several options:

From the partners:

Read carefully a list of projects proposed by our partners from this document. When choosing a project from a partner, consider not only a topic and the description of an idea, but also if the data is readily available, if a person is ready to spend enough time with a team, and if the complexity of the project is reasonable and aligns well with your expectations. You take full responsibility of working on this project, so be mindful. Also, when you sign up for a project from the partner make sure that your team does not exceed the limit on the number of teams that they are willing to supervise (this is indicated in the separate field). If you choose a project from one of the partners, we will connect you with the company/person who has proposed a project, so that you can get data and some guidance. Note, that extra time and care must be invested in communicating with a project owner and also meeting their expectations. Project owners will be invited to the final presentations and given an opportunity to influence the final assessment.

From Kaggle.com:

You can do a project by participating in one of the Kaggle.com competitions (e.g. Predict Future Sales) or working on some Kaggle dataset (e.g. solar power generation dataset). Kaggle challenges and the datasets are usually supported by ‘kernels’ where people document and publish their analysis code. You must declare all kernels that you use in your project and convince instructors (during presentations) that what you have done is different from the existing solutions and constitutes a sufficient amount of work worth 25 points.

Come up with a project yourself:

If neither option above is of interest to you, feel free to propose your own project, for example, you may decide to analyze some publicly available data (from here, here, here, here or here). Make sure there is a very explicit machine learning component in your project - you can consult with instructors if in doubt. If you already have a team, describe your project and the team in the project document. If you don’t have a team, you can advertise your project in Slack in the #project channel and try to attract team members.

Team formation (deadline: Oct 14)

After you have chosen a project (or came up with it yourself), you need to assemble a team or join the existing one. For that you can either advertise your project in the #project channel or add a comment to someone else's post, letting them know that you want to join. Remember that the team size should be between 2 and 4 people. If someone has already formed a team, get approval from them before adding your name to the existing team in the project document.

By the deadline on Oct 14 at 23:59 - Each team must add their project by specifying the title, project type (from partners, kaggle, or your own), project description, and team members into the project document. Follow the template in the beginning and check out our example project. If you plan to work on a project that was proposed by a partner, please, in the project type section specify the project identifier (e.g. based on P01 - Improving the quality of museums data). Also specify which day (Monday, Tuesday, or Wednesday) you want to make your intermediate and final presentation. This would determine which practice session instructors will be grading your presentations.

Intermediate presentation: Nov 14 - 16 (5 points out of 25 points)

General guidelines for intermediate presentations:

• At least one member from the team (can be more than one) will present. Other team members should attend as well, to support and help with QA.
• prepare a 5 minutes talk (with slides), if you go over time, we will cut you off;
• Add your google slides (and only!) to the corresponding google folder: here for Monday, here for Tuesday and here for Wednesday. Make sure to name your presentations TXX - title of the project e.g. “T02 - Teaching 1:10 scale cars to drive in a coty (P08)“.
• in the presentation make sure to introduce your team and project owner (if applicable);
• briefly describe the problem you are trying to solve (say why it needs to be solved);
• mention progress you have managed so far;
• tell about your blockers/problems;
• lastly, say few words about future steps (what are you going to accomplish for the final presentation).

Final presentation: Dec 12 - 14 (20 points out of 25 points)

• At least one member from the team (can be more than one) makes a presentation about the project. Others should attend as well, to support and help with QA.
• As intermediate presentations, the final presentation will be held offline (+ over Zoom if necessary) with two practice session leaders grading the presentations.
• You have 5 minutes to make the final presentation.
• Add your google slides (and only!) to the corresponding google folder: here for Monday, here for Tuesday and here for Wednesday. Make sure to name your presentations TXX - title of the project e.g. “T02 - Teaching 1:10 scale cars to drive in a coty (P08)“.
• in the presentation make sure to introduce your team and project owner (if applicable);
• briefly remind us the problem you are trying to solve (say why it needs to be solved);
• explain what was your approach to the problem (your methods);
• detail results you have obtained and how they match the initial expectations;
• Lastly, describe a few main lessons that you have learned while working on the project.

The order of teams and the schedule of final presentations is available here.

Grading

We will grade your presentations based on roughly the following criteria:

• amount and complexity of work performed (40%),
• quality of presentation (30%),
• degree to which you have completed the initial task (25%),
• being on time (5%).

Project grades are assigned to each team member equally unless there is a good reason to believe that some team members have done significantly less or more work than others. The intermediate presentation gives a maximum of 5 points to every team member, while the final presentation accounts for 20 points. Both intermediate and final presentations are to be done during practice sessions (exceptions could be Monday presentations). Each project will be graded by at least two instructors independently, the final grade will be their average. Getting at least 12 points for the project is a prerequisite for passing the course.

There is no written report for the project. The only thing we will ask you to submit is a link to your project GitHub and presentation (in PDF).

Short summary of what you have to do:

1. Choose a project a) from a partner b) from Kaggle.com or c) come up with a project idea yourself.
2. By October 14, form a team (2 - 4 people including you) using Slack channel #project or your contacts in the course and describe your project + team here: https://docs.google.com/document/d/1oVZW5aQXA_avYoKnk5AP51PI4z5ld5S1/edit#.
3. Deliver an intermediate presentation on Nov 14 - 16 during a chosen practice session (either on Monday, Tuesday or Wednesday)
4. Deliver a final presentation on Dec 12 - 14 during your chosen practice session