Dear Seminar participants,
I am pleased to announce two guest presentations with AI-related topics that are scheduled for Wednesday, Jan 25, 14:30 (s.t.) to 16:30 in Room WE5/04.004.
These two talks replace our Seminar meeting on that day. Please try to join the talks instead.
Best wishes, Christoph
Wed 25.1., 14:30 (s.t.) to 15:25, Room WE5/04.004
- Prof. Dr. Alexander Steen (Uni Greifswald)
- A standard translation for higher-order modal logics
- Abstract: Standard translations are a common tool for encoding modal logic formulas into classical logic in a truth-preserving way. Common standard translations include the mapping of propositional modal logic into unsorted first- order logic, and the mapping of first-order modal logic into many-sorted first- order logic. In contrast, encodings into higher-order logic (HOL) offer more flexibility and expressivity. In my talk, I will present ongoing work for a novel translation of higher-order (multi-)modal logics into HOL that supports both rigid and flexible constant/function symbols, different quantification semantics, local and global hypotheses. This work extends and partly simplifies previous work on semantic embeddings. A preliminary evaluation is presented.
Wed 25.1., 15:30 (s.t.) to 16:30, Room WE5/04.004
- Dr. Serge Autexier (Director of the Bremen Ambient Assisted Living Lab, DFKI Bremen)
- Security and Privacy by Design in the development of multi-center-based machine learning for personalized health and care: some best practices and practical challenges
- Abstract: Security and privacy are of utmost importance for systems processing personal health data. The talk presents the impact of security and privacy requirements on the research and development of a platform to train machine learning models on real patient data from multiple sources in a multi-centre setting. It discusses how they affect the design of the system architecture, the choice of methods as well as the whole the whole research and development process itself. The presentation is based on experiences from two European research projects concerned with developing machine learning based personalized risk predictions for cancer patients and COPD patients and sheds a light on the peculiarities of the used data, the target variables to predict and the mechanisms based on predictive models to support medical personnel.