Course syllabus
DAT625 Structured machine learning lp1 HT24 (7.5 hp)
Course is offered by the department of Computer Science and Engineering
Contact details
Lecturer and examiner: Simon Olsson (simonols@chalmers.se) -- English only please.
Teaching assistant: Selma Moqvist <mselma@chalmers.se>
In-person only and there will be no recording of in-class sessions.
Course purpose
The purpose of this course is to give a broad introduction of structured machine learning. Structured machine learning involves using knowledge about the data domain and the data generating process to formulate learning problems which are more data efficient, generalize better, and hopefully also scale to larger problems.
The course will cover the following three broad themes:
- Data Generating Processes --- How is our data generated and what do we know about it and the resulting domain of our data?
- Geometric Deep learning --- How can we impose useful restrictions on the models that we learn that are in line with what we know about our data?
- Generative models --- How can we mimic the data generating distribution?
Schedule
| Tuesdays 8.00-11:45 (Lectures) |
| Wednesdays 8.00-9:45 (Tutorials/labs) |
| Fridays 8.00-11:45 (Lectures) |
Do not get scared off by the long lectures, they are broken up into smaller sessions with question and in-class problem solving.
Lecture overview
| Day | Date | Time | Room | |||
| w36 | Tuesday | 2025-09-02 | 08:00 | 11:45 | TP-L24 | Introduction, course overview |
| w36 | Friday | 2025-09-05 | 08:00 | 11:45 | TP-L24 | Data generating processes, high-dimensional data |
| w37 | Tuesday | 2025-09-09 | 08:00 | 11:45 | TP-L24 | Geometric priors |
| w37 | Friday | 2025-09-12 | 08:00 | 11:45 | TP-L24 | Grids |
| w38 | Tuesday | 2025-09-16 | 08:00 | 11:45 | TP-L24 | Graphs and sets |
| w38 | Friday | 2025-09-19 | 08:00 | 11:45 | TP-L24 | Groups and Convolutions |
| w39 | Tuesday | 2025-09-23 | 08:00 | 11:45 | TP-L24 | Convolutions and group representations |
| w39 | Friday | 2025-09-26 | 08:00 | 11:45 | TP-L24 | Application domains: Molecules |
| w40 | Tuesday | 2025-09-30 | 08:00 | 11:45 | TP-L24 | Generative Models 1: Ambient and latent spaces, autoregressive, bayesian networks |
| w40 | Friday | 2025-10-03 | 08:00 | 11:45 | TP-L24 | Generative Models 2: Data manifold, denoising, Variational Auto Encoders |
| w41 | Tuesday | 2025-10-07 | 08:00 | 11:45 | TP-L24 | SELF-STUDY PROJECTS |
| w42 | Tuesday | 2025-10-14 | 08:00 | 11:45 | TP-L24 | Generative Models 3: Diffusion models |
| w42 | Friday | 2025-10-17 | 08:00 | 11:45 | TP-L24 | Generative Models 4: Normalizing flows |
| w43 | Tuesday | 2025-10-21 | 08:00 | 11:45 | TP-L24 | The Bitter Lesson; its impact and where do we go from here |
| w43 | Friday | 2025-10-24 | 08:00 | 11:45 | TP-L24 | Project Presentations |
Tutorial schedule
| Week | Date | Start time | End time | Room |
| w36 | 2025-09-03 | 08:00 | 09:45 | TP-L24 |
| w37 | 2025-09-10 | 08:00 | 09:45 | TP-L24 |
| w38 | 2025-09-17 | 08:00 | 09:45 | TP-L24 |
| w39 | 2025-09-24 | 08:00 | 09:45 | TP-L24 |
| w40 | 2025-10-01 | 08:00 | 09:45 | TP-L24 |
| w41 | 2025-10-08 | 08:00 | 09:45 | TP-L24 |
| w42 | 2025-10-15 | 08:00 | 09:45 | TP-L24 |
| w43 | 2025-10-22 | 08:00 | 09:45 | TP-L24 |
Assignment Schedule
| Release | Deadline | Final re-submission opportunity | Assessment | |
| Hand-in 1 | 02-Sep | 08-Sep | NA | Pass/fail |
| Hand-in 2 | 09-Sep | 15-Sep | NA | Pass/fail |
| Hand-in 3 | 16-Sep | 22-Sep | NA | Pass/fail |
| Project 1 | 22-Sep | 29-Sep | 15-Oct | Graded |
| Project 2 | 30-Sep | 07-Oct | 20-Oct | Graded |
| Project 3 | 08-Oct | 17-Oct | 22-Oct | Graded |
| Essay | 08-Sep | 20-Oct | 25-Oct | Pass/fail |
Course literature (to be updated)
Large parts of the course material is loose adaptations from Geometric Deep Learning summer schools.
Primary reference:
Bronstein, Bruna, Cohen, and Veličković: Geometric deep learning. (Free proto-book available: https://arxiv.org/abs/2104.13478)
Selected primary literature and lecture notes TBA
Extra literature:
Serre "Linear Representations of Finite Groups" (1977) https://link.springer.com/book/10.1007/978-1-4684-9458-7
Background references for repetition:
Deisenroth, Faisal, and Ong "Mathematics for Machine Learning" (2021) https://mml-book.github.io/book/mml-book.pdf
Shapira "Linear Algebra and Group Theory for Physicists and Engineers." (2019) https://link.springer.com/book/10.1007/978-3-030-17856-7
Petersen and Pedersen "The Matrix Cookbook" https://www.math.uwaterloo.ca/~hwolkowi/matrixcookbook.pdf
Git crash course: https://julienpascal.github.io/slides/intro_git/#/
Further math background Mathematical Foundations of Geometric Deep Learning
Practical extra information:
37 reasons why you neural network is not working (Blog post)
Course design
The course is based on four components
- Self study: reading and video lectures (where applicable)
- Interactive hybrid lectures
- Individual project work
- Peer-assessment
and follows a blended classroom structure.
Self study:
Before each week reading and video material is provided for preparation. Preparation is key for the success of the interactive lectures. In these lectures, we will be moving through concepts and problems, and solve them in teams. There will not be repetition of materials given in advance, we assume attendees have read and followed provided material in advance. The first lecture is an exception.
Lectures:
The in-class lectures are interactive and focus on problem solving in teams, interspersed with micro lectures, and discussions (tuesdays and wednesdays). Attendance is strongly advised.
Project work and assessment:
The course assessment involves four elements:
- Three individual projects each with equal weight towards the final grade. The projects focus on different aspects of structural learning but are problem-driven.
- Three assignments (pass/fail), and two of them need to be passed to pass the course
- A written project proposal should be submitted and approved (pass/fail) and a 5 minute pitch presentation.
- Peer assessment of the written proposals and questions after the pitch.
Learning objectives and syllabus
Examination form
To pass the course the following elements must be completed by the end of the course.
- three take home assignments, (pass/fail)
- three practical projects, (graded)
- one project proposal for a 6 month research project, max 1500 words (pass/fail) and 5 minute pitch.
- Peer review of project proposals and acting as opponent at presentation
Projects credit distribution:
To pass a project assignment you must:
- Pass unit tests
- Include a statement on the use of aids (Generative AI models, talking to friends, stack overflow etc.)
Your report will only be graded when the first steps are passed.
The written report can give up to 10 points distributed as follows:
- Clear introduction (1 points),
- clear motivation (1 points),
- appropriate references to external material (1 points)
- description of methods/aids used (1 points),
- presentation of results including high quality figures/illustrations (3.5 points)
- Discussion and conclusion (2.5 points)
Late submission policy: 0.5 points is deducted for every day of delay the project is submitted late, for a maximum of 8 days. There after a maximum of 6 points will be given for the assignment. Last submission deadline is given in the table above. Submissions after this deadline will not be graded and will result in a failing grade. Remember, exceptions are possible in exceptional cases; if you think you have one please reach out.
All aids are allowed throughout the course. However, the use of aids needs to be thoroughly documented in the project reports.
Course summary:
| Date | Details | Due |
|---|---|---|