Course organization

I have a problem, who should I contact?

If you have questions about course content or problems with coding assignments, please contact anyone from the course team by email or on Slack (invite).

There are too many platforms in this course, I am lost.
  1. The main course page is the MyCourses Aalto page. All important announcements will be released from MyCourses. There you can find all information about the course and links to other resources.
  2. We will be running our Python code (in the form of Jupyter notebooks) in Aalto Jupyter Hub. See Assignment instructions for more information.
  3. Third important platform is Slack, where teachers and students can easily communicate, post questions and discuss any related topics. You can join Slack with this invite.. Instructions how to join slack channels are here.
  4. We also use Zoom for online lectures. You can login to Zoom with your Aalto account (select sign in with SSO, your company domain is "aalto.zoom.us").
What does the "D" mean in the course name?

It means that this course can be included in a doctoral degree.

Lectures

Are the lectures recorded and shared?

Yes, lecture recordings will be made available afterwards.

Are there exercise sessions?

No, there will only be lectures twice per week. Instead of the exercise sessions, we have the jupyter notebooks that you can complete at your own pace which you'll also turn in as homework. You are encouraged to reach out in case you're unsure about anything in there!

What are the study materials for the course?

Assignments

How to use JupyterHub?

See Assignment instructions for more information.

Why can't I see the assignments?

The assignments will be released at the latest the day of the relevant lecture and they may not have been released yet.

Where can I find solutions for tasks after the deadline?

Solutions for most tasks will be found in "Assignments" section. In general, to see how a task was evaluated you need to fetch feedback in Jupyter Hub/Assignments tab, feedback will be released 1-2 workdays after the deadline. In the feedback file, you will see your points and solutions in "hidden tests" cells.

My coding exercise solution is almost correct, can I have partial points?

Depends. Ask TAs for evaluation.

ML project

Is the dataset selection limited in any way?

No, you may use any dataset you want. This includes your own research data, datasets from platforms like Kaggle.com or metrological ("weather") data (https://en.ilmatieteenlaitos.fi/download-observations-questions). 

Only using the same dataset, features, labels, models, losses,... as in one of the lectures/assignments is discouraged and penalized. This is very unlikely to happen by accident but you will not be penalized if the same data is shown after you have chosen it (after stage 1).

Will dataset X be a good fit for the project?

The goal of the student project is to let you practice the formulation (modelling) of some application of your interest as a ML problem. This formulation amounts to identifying the datapoints, their features and labels. Another goal of the project is to let you practice the application and comparison of different ML methods. The final performance of the models is not relevant at all for the evaluation of your project. 

It is very well possible that reaching "good" performance might be very difficult on some problems without more advanced methods. For example, image classification has seen an unprecendented improvement with the advance of convolutional neural networks. Not all image classification problems are like that, however, and you're encouraged to try and see for youself!

However, consider presenting a solution to a simplified version of the problem. For example, instead of classifying different hand-written digits, you can simply find a classifier of whether the image is a 0 or not.

Can I work on the project with a friend?

No, the ML project is an individual assignment, because you are meant to practice the whole process of applying ML to a problem.

How should the report look like?

The report must follow the structure described in the Section "ML project" (Deliverable for stage 1 - 3). Furthermore, example reports from last year's students will be shared (with the authors' consent).

What methods may I use for the project?

Due to peer-grading, only methods we have covered in this course are allowed. This will be a subset of scikit-learn methods and we'll publish a list of these in due time.

I'm worried about being penalized for peer grading inconsistency.

There is no need to worry if you do the peer-reviews properly and justify your grading  decision(s). Moreover, you will always have the possibility to review and appeal to the course staff if you consider any aspect of the grading unfair. We will then manually over-rule the automated peer-grading system if we find it appropriate.


Last modified: Tuesday, 1 February 2022, 10:20 AM