Welcome to the webpage of the course in Explainalbe Artificial Intelligence in the Life Sciences, M.Sc. in Life Science Informatics, Univeristy of Bonn
M-LSI-W-036, LC, 8 credits.
Course period: 28th Sept- 8th Oct, 10AM-5PM
Room: U1.105 (b-it building, Friedrich-Hirzebruch-Allee 6)
Lab teaching assistant: Julia Teja Urrutia (tejaurrutia [at] bit.uni-bonn.de)
The course is mainly divided into two parts:
Theory (morning)
Exercises (afternoon)
Evaluation: presentation of the final exercise done during the lab hours
After the end of the course, an anonymous questionnaire will be made available to ask for feedback on the course.
Knowledge
The students will learn the foundations of explainable artificial intelligence, along with specific tools and techniques to explain black-box machine learning models with applications in the life sciences, presented during theoretical lectures.
Skills
The students will learn how to implement and use the main explainable artificial intelligence algorithms for life sciences using Python during practical hands-on coding sessions.
Competences
The students will be able to critically analyze problems that require explainability and choose the right explanation strategy according to the task under consideration and the scientific requirements.
Contents:
Theory and foundations of explainable artificial intelligence
Self-explainable models
Feature attribution methods
Rule-based methods
Example-based methods
Practical hands-on sessions with relevant use cases
Sept 28:
Theory: Course info [slides], Theory and Foundations of XAI [slides], Self-explainable models [slides]
Lab: Chemical data handling and self-explainable models [notebooks: exercise, solution]
Sept 29:
Theory: Feature attribution methods [slides]
Lab: Feature attribution: LIME [notebooks: exercise, solution]
Sept 30:
Oct 1:
Theory: Example-based methods [slides]
Lab: Feature attribution: Neural networks with PyTorch and Integrated Gradients [notebooks: exercise, solution]
Oct 2:
Theory: Explaining graph neural networks [slides]
Lab: GNNs with PyTorch Geometric and GNNExplainer [notebooks: exercise, solution]
Oct 5-7:
Lab: final exercise [guidelines: slides]
Oct 8:
Evaluation