CIT423007 - Trustworthy AI for Medicine
INF3151 - Introduction to Machine Learning (“Grundlagen des maschinellen Lernens”)

CIT423007 - Trustworthy AI for Medicine

Lecturers

Prof. Zeynep Akata, Dr. Georgios Kaissis, Prof. Daniel Rückert

Content

This module covers the core concepts of trustworthy AI in medicine:

  • Technical robustness and safety: reliable AI design, uncertainty quantification, robustness and verification, safety mechanisms
  • Transparency and explainability: XAI methods (e.g., LIME, SHAP), traceability and auditability, transparency of systems
  • Ethical, legal, and social implications: fairness and bias mitigation, privacy and data governance, human agency and oversight, accountability and responsibility

The first part of the lecture is given by the EML Munich group. These lectures introduce the technical foundations of modern large-scale AI (foundation models, adaptation, multimodal alignment, generative models) and connect them to explainability and trustworthiness.


This is a Master's level course with 4 semester weekly hours, taught in English.

Organisation

The course runs in the winter semester 2026/27 (14.10.2026 - 05.02.2027):

  • Lectures: Wednesdays, 10:00-12:00, Seminarraum 5901.02.013 (02.5901.013)
  • Tutorials: Fridays, 10:00-12:00, Seminarraum 5901.01.013 (01.5901.013)

The schedule is as follows:

  • Week 1 (12.-16.10.): Introduction and fundamentals (Prof. Zeynep Akata)
  • Week 2 (19.-23.10.): Large-scale models and adaptation (Yiran Huang)
  • Week 3 (26.-30.10.): Parameter efficient fine-tuning (Dr. Massimo Bini)
  • Week 4 (2.-6.11.): Multimodal alignment (Simon Roschmann)
  • Week 5 (9.-13.11.): Generative models (Dr. Maria Bravo)
  • Week 6 (16.-20.11.): Mechanistic interpretability (Dr. Kirill Bykov)
  • Week 7 (23.-27.11.): Uncertainty taxonomy and Bayesian neural networks
  • Week 8 (30.11.-4.12.): Calibration, conformal prediction, and OOD detection
  • Week 9 (7.-11.12.): Privacy-preserving machine learning
  • Week 10 (14.-18.12.): Federated learning and membership inference attacks
  • Week 11 (11.-15.1.): Active learning with differential privacy and data attribution
  • Week 12 (18.-22.1.): Memorisation, bias, and fairness
  • Week 13 (25.-29.1.): Advanced topics in memorisation
  • Week 14 (1.-5.2.): Summary

Registration and requirements

The module is required for the M.Sc. AI in Biomedicine, whose students have priority. Enrollment is limited to 50 participants. Solid programming skills and practical experience with deep learning frameworks are essential.

Please refer to the TUM course page for up-to-date information.