Informed Machine Learning

Daniel Schulz is one of the managing directors of the Fraunhofer Cluster of Excellence Cognitive Internet Technologies CCIT, where he is responsible for the Fraunhofer Technology Hub Machine Learning and works on implementable technology solutions for the edge-cloud continuum. His main research focuses on informed machine learning techniques that not only learn from data but can also utilize existing knowledge and models. In addition, Daniel Schulz represents the Fraunhofer Institute for Intelligent Analysis and Information Systems (IAIS) at the Scientific and Technical Council of the Fraunhofer Society. He studied Geosciences at the Universities of Cologne, Bonn and Gothenburg, and has today 15+ years of experience as a senior data scientist in industry and public funded projects in various industries and research fields.Christian Bauckhage is a professor of computer science (intelligent learning systems) at the University of Bonn, lead scientist for machine learning at Fraunhofer IAIS, and one of the directors of the Lamarr Institute for Machine Learning and Artificial Intelligence. He has 20+ years of experience as a data scientist in industry and academia and (co)authored numerous publications on pattern recognition, data mining, and machine learning. His current research focuses on informed machine learning techniques that integrate knowledge- and data-driven methods. Practical applications of his work can be found in fields as diverse as physics, agriculture, or business analytics. As an expert on applied AI, he frequently consults private and public institutions regarding the design and deployment of intelligent systems.

• Preface.- 1. Introduction and Overview.- Part I. Digital Twins.-  2 Optimizing Cooling System Operations with Informed ML and a Digital Twin.- 3. AITwin - A Uniform Digital Twin Interface for Artificial Intelligence Applications.- Part II. Optimization.- 4. A Regression-based Predictive Model Hierarchy for Nonwoven Tensile Strength Inference.- 5. Machine Learning for Optimizing the Homogeneity of Spunbond Nonwovens.- 6. Bayesian Inference for Fatigue Strength Estimation.- 7. Incorporating Shape Knowledge into Regression Models.- Part III Neural Networks.- 8. Predicting Properties of Oxide Glasses Using Informed Neural Networks.- 9. Graph Neural Networks for Predicting Side Effects and New Indications of Drugs Using Electronic Health Records.- 10. On the Interplay of Subset Selection and Informed Graph Neural Networks.- 11. Informed Machine Learning Aspects for the Multi-Agent Neural Rewriter.- Part IV. Hybrid Methods.- 12. Training Support Vector Machines by Solving Differential Equations.- 13. Informed Machine Learning to Maximize Robustness and Computational Performance of Linear Solvers.- 14. Anomaly Detection in Multivariate Time Series Using Uncertainty Estimation.