Advanced Topics in AI and Robotics

Abstract

Complex machine learning models are applied more and more. Their black box nature has elicited research to make them more explainable. However, most explainability methods cannot provide insight beyond the given data, requiring additional information about the context. In this talk, Beckh will argue that harnessing prior knowledge can improve explanations, e.g. to make them more accessible, and provide three main approaches how. She will present the existing taxonomy of informed machine learning as a starting point, provide examples for each knowledge-driven explainable ML approach and highlight open challenges and research directions.

Short Bio

Katharina Beckh is a research scientist at Fraunhofer IAIS in a natural language understanding team. She received her MSc degree in human-computer interaction from the Julius Maximilian University of Wuerzburg in 2019. Since then, she has worked on and managed data science projects in several domains including health care, electrical engineering and renewable energy. Her research interest is in explainable natural language processing with a focus on evaluating, utilizing and improving explanations.

Papers relevant for the talk

1. Beckh, K., Müller, S., Jakobs, M., Toborek, V., Tan, H., Fischer, R., Welke, P., Houben, S. and von Rueden, L., SoK: Harnessing Prior Knowledge for Explainable Machine Learning: An Overview. In First IEEE Conference on Secure and Trustworthy Machine Learning. (https://openreview.net/forum?id=1KE7TlU4bOt)

2. L. von Rueden et al., "Informed Machine Learning – A Taxonomy and Survey of Integrating Prior Knowledge into Learning Systems," in IEEE Transactions on Knowledge and Data Engineering, vol. 35, no. 1, pp. 614-633, 1 Jan. 2023, doi: 10.1109/TKDE.2021.3079836. (https://ieeexplore.ieee.org/abstract/document/9429985)

3. Sokol, K., Flach, P. One Explanation Does Not Fit All. Künstl Intell 34, 235–250 (2020). (https://link.springer.com/article/10.1007/s13218-020-00637-y

The lecture will be held in English and is aimed at students and staff of the H-BRS. Interested parties are cordially invited.