Seminar on Machine Learning in Health
Organizers
Dr. Maria Kalweit
Prof. Dr. Joschka Boedecker
Co-organizers
Dr. Gabriel Kalweit, Mehdi Naouar, Yannick Vogt, Lisa Graf
Description
The format is the following:
Introductory lecture: We will discuss general aspects of scientific presentations in the beginning and define the format of the seminar.
Phase I: Each participant will become an expert in one methodological core area by reading one or multiple research papers, which we will assign after the students express their interest. Then, the students will form teams in the respective core areas and discuss advantages and disadvantages of their papers together.
Phase II: The students will again form groups to discuss specific medical use cases in light of the methodological core areas. Each group in Phase II will consist of one expert from each core area.
At the end of the semester, there will be a (potentially open-door) mini-symposium, in which the students will present their findings in both a prepared talk and a poster session.
Course information
| Details: |
Course number: 11LE13S-7341-M
Attendees: 12
NEXUS Lab on the ground floor of the new IMBIT building (Georges-Köhler-Allee 201). The meetings are in person. |
| Course Schedule: |
Introduction: Monday, October 14th, 2024, 10:00 - 11:30 (s.t.) |
| Requirements: | Prior knowledge in Machine Learning and Deep Learning. |
| Selection Procedure: | Please register via HISInOne. Candidates are prioritised if they either are present at the introduction event or write a short motivational mail to Maria Kalweit until Wednesday, October 16th 23:59am. The subject line should include [ML in Health]. After you have been confirmed, you will then bid on a topic as discussed in the introduction event. |
| Resources: | Poster guideline as pptx or pdf |
| ILIAS: | TBA |
Timeline
All deadlines are due at 11:59 PM GMT+2 (These are hard deadlines!) and all meetings are in person and mandatory. Please only register if you can be there in person in every meeting.
| Date | Done | Comment | |
| Announcement paper list: | Friday, October 11th | ✓ | Topics are listed below |
| Paper Voting Deadline: | Thursday, October 31st | Assignments will be announced Monday, November 4th via email! | |
| First Meeting with Supervisors: | Friday, November 29th | Please send a pdf of the poster draft via email to Maria Kalweit until Monday, November 25th. We will meet in IMBIT, room 42 (first floor, beside the kitchen). First, you will get an introduction into your core area. Then, you will present your poster in a small round. Please prepare, that is a dry-run of the symposium to get feedback (the draft will not be graded). | |
| Final Poster Deadline: | Monday, December 9th | Please send a pdf of the poster via email until Monday, December 9th. | |
| Announcement Medical Use Cases | Monday, December 9th | We will announce the medical use cases via email. | |
| Second Meeting with Supervisors: | Friday, January 24th |
|
Please send a pdf or pptx of the presentation draft until Monday, January 20th. We will meet in IMBIT, room 42 (first floor, beside the kitchen). You can give a test-talk and receive feedback (the draft will not be graded). |
| Final Presentation Deadline: | Sunday, February 2nd | ||
| Block Seminar [Format: Symposium]: | Friday, February 7th |
1st Meeting with Supervisors
We will meet in IMBIT, room 42 (1st floor behind the kitchen) to discuss your poster drafts and general questions.
Category A: Explainability 9:15 - 10:15
Category B: Efficiency 10:15 - 11:15
Category C: Accessible AI 13:00 - 14:00
Category D: Trustworthy AI 14:00 - 15:00
2nd Meeting with Supervisors
We will meet in IMBIT, room 42 (1st floor behind the kitchen) to discuss your presentation drafts and general questions.
Medical Use Case 1: TBA 9:15 - 10:15
Medical Use Case 2: TBA 10:15 - 11:15
Medical Use Case 3: TBA 13:00 - 14:00
Presentations
Once the posters are created, we will work on specific medical use cases. Then, you will work in a team with experts of the different core areas (category A-D) to create an oral presentation with slides (we will have 3 oral presentations).
Posters
The poster assignment will be announced via email. Everyone creates his or her own poster about his or her assigned paper, which he or she will then present at the mini-symposium in the poster session (we will have 12 posters in total). In your core area group (category A - D), we will shortly go over the corresponding topic and talk about all posters on November 29, so that all of you get an overview over the core area and to get feedback on your paper from the supervisors and the other students.
Voting
If you got accepted to the seminar, please send your priorities in an email to Maria Kalweit. The email should have the following form:
Subject line: [ML in Health – Paper Voting]
Text:
[Your Priority list]
The list should have decreasing order (highest priority at the top, lowest at the bottom) and a line break after each topic. Please list all papers (12 entries, one line per paper ID). For example:
B.1
B.2
A.1
B.3
...
Please send the email until Thursday, October 31 11:59PM the latest. Any emails received after this deadline cannot be considered in the voting process. We will do our best to accommodate your preferences and ensure that you are assigned papers that align with your interests.
Topics
Category A: Explainability
- A.1: Predict and Interpret Health Risk Using Ehr Through Typical Patients
- A.2: Explainable Histopathology Image Classification with Self-organizing Maps: A Granular Computing Perspective
- A.3: Recent advances in interpretable machine learning using structure-based protein representations
Category B: Efficiency
- B.1: HyperFast: Instant Classification for Tabular Data
- B.2: Scaling Vision Transformers to Gigapixel Images via Hierarchical Self-Supervised Learning
- B.3: Geometric epitope and paratope prediction
Category C: Accessible AI
- C.1: Zero shot health trajectory prediction using transformer
- C.2: A multimodal generative AI copilot for human pathology
- C.3: AREA-AFFINITY: A Web Server for Machine Learning-Based Prediction of Protein–Protein and Antibody–Protein Antigen Binding Affinities
Category D: Trustworthy AI
- D.1: Navigating the EU AI Act: implications for regulated digital medical products
- D.2: Mitigating Neural Network Overconfidence with Logit Normalization
- D.3: Robust deep learning–based protein sequence design using ProteinMPNN