@inproceedings{ZellerWaschulzikCarlanetal.2024, author = {Zeller, Marc and Waschulzik, Thomas and Carlan, Carmen and Serahlazau, Marat and Bahlmann, Claus and Wu, Zhiliang and Spieckermann, Sigurd and Krompass, Denis and Geerkens, Simon and Sieberichs, Christian and Kirchheim, Konstantin and {\"O}zen, Batu Kaan and Robles, Lucia Diez}, title = {Continuous Development and Safety Assurance Pipeline for ML-Based Systems in the Railway Domain}, series = {Computer Safety, Reliability, and Security. SAFECOMP 2024 Workshops. DECSoS, SASSUR, TOASTS, and WAISE, Florence, Italy, September 17, 2024, Proceedings. Lecture Notes in Computer Science, vol 14989}, booktitle = {Computer Safety, Reliability, and Security. SAFECOMP 2024 Workshops. DECSoS, SASSUR, TOASTS, and WAISE, Florence, Italy, September 17, 2024, Proceedings. Lecture Notes in Computer Science, vol 14989}, editor = {Ceccarelli, Andrea and Trapp, Mario and Bondavalli, Andrea and Schoitsch, Erwin and Gallina, Barbara and Bitsch, Friedemann}, publisher = {Springer Nature}, address = {Cham}, isbn = {9783031687372}, issn = {0302-9743}, doi = {10.1007/978-3-031-68738-9_36}, pages = {446 -- 459}, year = {2024}, subject = {Maschinelles Lernen}, language = {en} } @unpublished{GeerkensSieberichsBraunetal.2023, author = {Geerkens, Simon and Sieberichs, Christian and Braun, Alexander and Waschulzik, Thomas}, title = {QI2 -- an Interactive Tool for Data Quality Assurance}, publisher = {arXiv}, year = {2023}, abstract = {The importance of high data quality is increasing with the growing impact and distribution of ML systems and big data. Also the planned AI Act from the European commission defines challenging legal requirements for data quality especially for the market introduction of safety relevant ML systems. In this paper we introduce a novel approach that supports the data quality assurance process of multiple data quality aspects. This approach enables the verification of quantitative data quality requirements. The concept and benefits are introduced and explained on small example data sets. How the method is applied is demonstrated on the well known MNIST data set based an handwritten digits.}, language = {en} } @unpublished{SieberichsGeerkensBraunetal.2023, author = {Sieberichs, Christian and Geerkens, Simon and Braun, Alexander and Waschulzik, Thomas}, title = {ECS -- an Interactive Tool for Data Quality Assurance}, publisher = {arXiv}, doi = {10.48550/arXiv.2307.04368}, year = {2023}, abstract = {With the increasing capabilities of machine learning systems and their potential use in safety-critical systems, ensuring high-quality data is becoming increasingly important. In this paper we present a novel approach for the assurance of data quality. For this purpose, the mathematical basics are first discussed and the approach is presented using multiple examples. This results in the detection of data points with potentially harmful properties for the use in safety-critical systems.}, language = {en} } @article{GeerkensSieberichsBraunetal.2024, author = {Geerkens, Simon and Sieberichs, Christian and Braun, Alexander and Waschulzik, Thomas}, title = {QI²: an interactive tool for data quality assurance}, series = {AI and Ethics}, volume = {4}, journal = {AI and Ethics}, publisher = {Springer Nature}, issn = {2730-5961}, doi = {10.1007/s43681-023-00390-6}, url = {http://nbn-resolving.de/urn:nbn:de:hbz:due62-opus-42945}, pages = {141 -- 149}, year = {2024}, abstract = {The importance of high data quality is increasing with the growing impact and distribution of ML systems and big data. Also, the planned AI Act from the European commission defines challenging legal requirements for data quality especially for the market introduction of safety relevant ML systems. In this paper, we introduce a novel approach that supports the data quality assurance process of multiple data quality aspects. This approach enables the verification of quantitative data quality requirements. The concept and benefits are introduced and explained on small example data sets. How the method is applied is demonstrated on the well-known MNIST data set based an handwritten digits.}, language = {en} } @article{SieberichsGeerkensBraunetal.2024, author = {Sieberichs, Christian and Geerkens, Simon and Braun, Alexander and Waschulzik, Thomas}, title = {ECS: an interactive tool for data quality assurance}, series = {AI and Ethics}, volume = {4}, journal = {AI and Ethics}, publisher = {Springer Nature}, issn = {2730-5961}, doi = {10.1007/s43681-023-00393-3}, url = {http://nbn-resolving.de/urn:nbn:de:hbz:due62-opus-42908}, pages = {131 -- 139}, year = {2024}, abstract = {With the increasing capabilities of machine learning systems and their potential use in safety-critical systems, ensuring high-quality data is becoming increasingly important. In this paper, we present a novel approach for the assurance of data quality. For this purpose, the mathematical basics are first discussed and the approach is presented using multiple examples. This results in the detection of data points with potentially harmful properties for the use in safety-critical systems.}, language = {en} } @techreport{GeerkensSieberichs2026, author = {Geerkens, Simon and Sieberichs, Christian}, title = {safe.trAIn - sichere Ki am Beispiel fahrerloser Regionalzug: Abschlussbericht des Projektes safe.trAIn - sichere Ki am Beispiel fahrerloser Regionalzug. Partner: Hochschule D{\"u}sseldorf}, publisher = {Technische Informationsbibliothek}, address = {Hannover}, doi = {10.34657/29998}, url = {http://nbn-resolving.de/urn:nbn:de:hbz:due62-opus-60980}, pages = {17}, year = {2026}, abstract = {Mit dem EU AI Act wurden KI-Systeme mit sicherheitskritischen Funktionen als Hochrisiko-Anwendungen eingestuft. Sie m{\"u}ssen hohen Anforderungen an Datenqualit{\"a}t, Robustheit, Transparenz und Konformit{\"a}tsbewertung gen{\"u}gen. Forschungsarbeiten zur Erkl{\"a}rbarkeit neuronaler Netze (z. B. LRP, CAM) und zur Auswirkung optischer Eigenschaften auf KI-Systeme bilden die Basis. Zudem wurde erkannt, dass fehlerhafte Daten und optische Abbildungsfehler die Sicherheit stark beeintr{\"a}chtigen k{\"o}nnen. Ziel des Projekts safe.trAIn war es, die Voraussetzungen f{\"u}r einen sicheren vollautomatisierten Zugbetrieb zu schaffen. Im Fokus stand die Absicherung KI-basierter Verfahren zur Hinderniserkennung und deren Integration in Sicherheitsnachweise. Die Hochschule D{\"u}sseldorf (HSD) {\"u}bernahm zentrale Aufgaben bei der Entwicklung und Weiterentwicklung von Methoden zur Daten- und KI-Absicherung, insbesondere den Queens-Methoden (QI² und ECS). Methoden: - Entwicklung und Anwendung der Queens-Methoden zur Bewertung von Datenqualit{\"a}t und Modellrobustheit - Analyse von Unsicherheiten in KI-gest{\"u}tzter Objekterkennung und deren Propagation - Entwicklung physikalisch realistischer Optikmodelle zur Untersuchung sensorischer Einfl{\"u}sse - Validierung im virtuellen Testfeld (GoA4-System, driverless operation) - Integration in Sicherheitsarchitekturen (RAMS- und GSN-Methodik) sowie Begutachtung durch T{\"U}V. Ergebnisse: - Erfolgreiche Weiterentwicklung der Queens-Methoden (QI², ECS) f{\"u}r Daten- und Modellbewertung - Entwicklung von Verfahren zur Erkl{\"a}rbarkeit (Kombination QI² mit LRP) - Neue Ans{\"a}tze zur datenqualit{\"a}tsbasierten Fehlererkennung ohne Training - Optische Modellierungen und verbesserte Kamerakalibrierung - Beitrag zur Standardisierung (DIN DKE SPEC 99002) - Mehrere wissenschaftliche Publikationen und Betreuung von Promotionen und Abschlussarbeiten. Die entwickelten Methoden erm{\"o}glichen eine systematische Absicherung von KI-Systemen im Bahnumfeld. Anwendungsm{\"o}glichkeiten bestehen in der Integration in Zulassungsprozesse, bei der Datenkuratierung, in der Sensorikgestaltung sowie in weiteren sicherheitskritischen Bereichen. Dar{\"u}ber hinaus sind die Methoden anschlussf{\"a}hig f{\"u}r andere Forschungsfelder (z. B. Transformer-Netzwerke, chaostheoretische Analysen). Damit wurde eine technologische Grundlage f{\"u}r den sicheren Betrieb fahrerloser Regionalz{\"u}ge geschaffen.}, subject = {K{\"u}nstliche Intelligenz}, language = {de} }