KI 2022: Advances in Artificial Intelligence: 45th German Conference on AI, Trier, Germany, September 19–23, 2022, Proceedings (Lecture Notes in Computer Science, 13404)

This book constitutes the refereed proceedings of the 45 th German Conference on Artificial Intelligence, KI 2022, held in September 2022. The 12 full and 5 short papers were carefully reviewed and selected from 51 submissions. Additionally, five abstracts of invited talks are included. As well-established annual conference series KI is dedicated to research on theory and applications across all methods and topic areas of AI research. Due to COVID-19 the conference was held virtually. The chapter "Dynamically Self-Adjusting Gaussian Processes for Data Stream Modelling" is available open access under a Creative Commons Attribution 4.0 International License via link.springer.com. Preface Organization Abstracts of Invited Talks Prospects for Using Context to Integrate Reasoning and Learning Representation and Quantification of Uncertainty in Machine Learning The First Rule of AI: Hard Things are Easy, Easy Things are Hard Federated Learning: Promises, Opportunities and Security Challenges AI in Robotics and AI in Finance: Challenges, Contributions, and Discussion Contents An Implementation of Nonmonotonic Reasoning with System W 1 Introduction 2 Background on Conditional Logic 3 System W 4 Implementation and System Walkthrough 5 Conclusions and Further Work References Leveraging Implicit Gaze-Based User Feedback for Interactive Machine Learning 1 Introduction 2 Background 2.1 Confusion Detection 2.2 Leveraging Emotion Detection for ML 3 Method 3.1 Data Collection 3.2 Disagreement Detection Model 3.3 Application in IML 3.4 Limitations 4 Conclusion References The Randomness of Input Data Spaces is an A Priori Predictor for Generalization 1 Introduction 2 Randomness of Data Spaces 3 Experiments and Discussion 3.1 Synthetic Classification Problems with Known Decision Boundaries 3.2 Natural Data with Unknown Decision Boundaries 3.3 Studying Randomness of Input Data Spaces Without Randomization 4 Related Work 5 Conclusion 6 Limitations and Future Work References Communicating Safety of Planned Paths via Optimally-Simple Explanations 1 Introduction 2 Previous Work 3 Notation and Preliminaries 3.1 Road Map Domain 3.2 Form of Explanation 3.3 Constraint Parameterization and Subset Checking 3.4 Largest Possible Hulls 4 Optimal Explanation 4.1 Formal Requirements 4.2 Reduced Search Space sub 4.3 Solution from sub 5 Simulation 6 Conclusion and Future Work References Assessing the Performance Gain on Retail Article Categorization at the Expense of Explainability and Resource Efficiency 1 Introduction 2 Related Work 3 Experiments 3.1 Data 3.2 Methods 3.3 Evaluation 4 Discussion 5 Conclusion and Future Work References Enabling Supervised Machine Learning Through Data Pooling: A Case Study with Small and Medium-Sized Enterprises in the Service Industry 1 Introduction 2 Current State of Research and Application 3 Methods 4 Results 5 Discussion and Conclusion References Unsupervised Alignment of Distributional Word Embeddings 1 Introduction 2 Motivation and Related Work 2.1 Point-Based Word Embeddings 2.2 Probabilistic Embedding 2.3 Minimally Supervised Alignment of Word Embeddings 2.4 Fully Unsupervised Alignment of Word Embeddings 3 Approach 3.1 Problem Formulation 3.2 Orthogonal Procrustes 3.3 Wasserstein Procrustes 3.4 Wasserstein Procrustes for Gaussian Embedding 4 Experiments 4.1 Data Generation 4.2 Experimental Setup 4.3 Discussion 4.4 Data Generation 4.5 Experimental Setup 5 Conclusion References NeuralPDE: Modelling Dynamical Systems from Data 1 Introduction 2 Related Work 3 Task 4 Neural PDE 4.1 Method of Lines 4.2 NeuralPDEs 5 Data 6 Experiments 6.1 NeuralPDE Architecture 6.2 Comparison Models 6.3 Training 7 Results 8 Discussion 9 Conclusion References Deep Neural Networks for Geometric Shape Deformation 1 Introduction 2 Shape Deformation Modeling 2.1 Preparation of the Dataset 2.2 Implementation Results 3 Conclusion and Future Work References Dynamically Self-adjusting Gaussian Processes for Data Stream Modelling 1 Introduction 2 Related Work 3 ETKA 4 Experimental Setup 4.1 Simulated Data 4.2 Real-World Data 4.3 Evaluation 5 Experimental Results 5.1 Simulated Data 5.2 Real-World Data 5.3 Discussion 6 Conclusion References Optimal Fixed-Premise Repairs of EL TBoxes*-10pt 1 Introduction 2 Preliminaries 3 Generalized-Conclusion Repairs of EL TBoxes 4 Fixed-Premise Repairs of EL TBoxes 5 Complexity of Maximally Strong sub-Weakenings 6 Conclusion References Health and Habit: An Agent-based Approach 1 Introduction 2 Basics of Behavioural Theory and Agent Modeling 2.1 Health Action Process Approach: From Wanting to Doing 2.2 Social Learning: If You Can Do It, Maybe I Can Do It Too 2.3 Forming Habits 2.4 Beliefs-Desires-Intentions Architecture: Assembling Cognition 3 Bringing Everything Together: Concept Proposal 3.1 Translating HAPA to BDI 3.2 Introducing Influence 3.3 Forming Intentions and Habits 4 Discussion and Use Cases 4.1 Proof of Concept: Exercise in Heart Failure Patients 4.2 Related Works: Behaviour, Social Learning and Habit in ABMs 5 Conclusion References Knowledge Graph Embeddings with Ontologies: Reification for Representing Arbitrary Relations 1 Introduction 2 Preliminaries 2.1 Ortholattice and Orthologic 2.2 Background Logic 3 Cone Embedding 4 Distributive Embedding 5 Related Work 6 Conclusions and Outlook References Solving the Traveling Salesperson Problem with Precedence Constraints by Deep Reinforcement Learning 1 Introduction 2 Related Work 3 Problem Setting 4 Methodology 4.1 Encoder 4.2 Decoder 5 Experiments 5.1 Training and Datasets 5.2 Model Adaptations and Sparsification 5.3 Evaluation and Baselines 6 Results 7 Conclusion and Future Work A Appendix References HanKA: Enriched Knowledge Used by an Adaptive Cooking Assistant 1 Introduction 2 Methodology 2.1 Underlying Knowledge Representation 2.2 Configurations per Cooking Session 2.3 Adaptation While Cooking 3 System Description 3.1 Device Coordination 3.2 Frontend Coordination 3.3 Knowledge Representation 3.4 Monitoring and Execution 3.5 Planner 4 Evaluation 5 Conclusion References Automated Kantian Ethics: A Faithful Implementation 1 Introduction 2 The Need for Faithful, Explainable Automated Ethics 3 Automated Kantian Ethics 4 Details 4.1 Formalizing the Categorical Imperative in DDL 4.2 Isabelle/HOL Implementation 4.3 Testing Framework 5 Future Work 6 Related Work 7 Conclusion A Why Automate Kantian Ethics A.1 Consequentialism A.2 Virtue Ethics A.3 Kantian Ethics B Relevant Features of the Embedding of DDL in HOL C Additional Implementation Details D Experimental Figures E Additional Tests References PEBAM: A Profile-Based Evaluation Method for Bias Assessment on Mixed Datasets 1 Introduction 2 Theoretical Background 3 Methodology 3.1 Profile-Selection Based on Clustering 3.2 Evaluation of Clustering Methods for Profile Selection 3.3 Profile-Based Evaluation of a Given Classifier 4 Experiments and Results 4.1 Evaluation of Clustering for Profile Selection 4.2 Profile-Based Evaluation of Bias 5 Conclusion A Profiles on the German Credit Dataset References Author Index