Logo des Repositoriums

Auflistung nach Schlagwort "ML"

1 - 7 von 7
  • Konferenzbeitrag
    An Anthropomorphic Approach to establish an Additional Layer of Trustworthiness of an AI Pilot
    (Software Engineering 2022 Workshops, 2022) Regli, Christoph; Annighoefer, Björn
    AI algorithms promise solutions for situations where conventional, rule-based algorithms reach their limits. They perform in complex problems yet unknown at design time, and highly efficient functions can be implemented without having to develop a precise algorithm for the problem at hand. Well-tried applications show the AI’s ability to learn from new data, extrapolate on unseen data, and adapt to a changing environment — a situation encountered in fl ight operations. In aviation, however, certifi cation regulations impede the implementation of non-deterministic or probabilistic algorithms that adapt their behaviour with increasing experience. Regulatory initiatives aim at defining new development standards in a bottom-up approach, where the suitability and the integrity of the training data shall be addressed during the development process, increasing trustworthiness in eff ect. Methods to establish explainability and traceability of decisions made by AI algorithms are still under development, intending to reach the required level of trustworthiness. This paper outlines an approach to an independent, anthropomorphic software assurance for AI/ML systems as an additional layer of trustworthiness, encompassing top-down black-box testing while relying on a well-established regulatory framework.
  • Konferenzbeitrag
    HollerithEnergyML: A Prototype of a Machine Learning Energy Consumption Recommender System
    (INFORMATIK 2024, 2024) Zanger, Michael; Schulz, Alexander; Grodmeier, Lukas; Agaj, Dion; Schindler, Rafael; Weiss, Lukas; Möhring, Michael
    Energy consumption aspects of machine learning classifiers are important for research and practice as well. Due to sparse research in this area, a prototype of a recommender system was developed to provide energy consumption recommendations of different possible classifiers. The prototype is demonstrated as well as discussed and future research points are derived.
  • Konferenzbeitrag
    Künstliche Intelligenz in den Fingerspitzen
    (Workshops der 21. Fachtagung Bildungstechnologien (DELFI), 2023) Witt, Clemens
    In diesem Beitrag werden zwei kollaborative Lernspiele für Multitouchdisplays im Themenfeld „Künstliche Intelligenz“ vorgestellt. Diese erlauben Schüler:innen der Sekundarstufe I eine eigenständige und spielerische Auseinandersetzung mit den Problemlöseprozessen des maschi-nellen Lernens. Durch ihre systemunabhängige Implementierung können sie auf beliebigen Endgeräten genutzt und in vielfältigen Spielszenarien eingesetzt werden.
  • Konferenzbeitrag
    Machine Learning in Glass Bottle Printing Quality Control: A Collaboration with a Medium-Sized Industrial Partner
    (INFORMATIK 2024, 2024) Bundscherer, Maximilian; Schmitt, Thomas H.; Bocklet, Tobias
    In cooperation with a medium-sized industrial partner, we developed and evaluated two ML-based approaches for quality control in glass bottle printing. Our first approach utilized various filters to suppress reflections, image quality metrics for image comparison, and supervised classification models, resulting in an accuracy of 84%. We used the ORB algorithm for image alignment and to estimate print rotations, which may indicate manufacturing anomalies. In our second approach, we fine-tuned pre-trained CNN models, which resulted in an accuracy of 87%. Utilizing Grad-CAM, an Explainable AI method, we localized and visualized frequently defective bottle print regions without explicitly training our models for this use case. These insights can be used to optimize the actual manufacturing process beyond classification. This paper also describes our general approach and the challenges we encountered in practice with data collection during ongoing production, unsupervised preselection, and labeling.
  • Textdokument
    A Platform Framework for the Adoption and Operation of ML-based Smart Services in the Data Ecosystem of Smart Living
    (INFORMATIK 2022, 2022) Kortum,Henrik; Kohl,Tobias; Hubertus,Dominik; Hinz,Oliver; Thomas,Oliver
    Smart services utilizing machine learning (ML) take a more and more important position in our daily lives. As a result, the need for a large smart living data ecosystem has emerged that links the most diverse areas of life with each other. This ecosystem is characterized by a multitude of different actors, a heterogeneous system, product and service landscape as well as high data protection requirements. To provide real added value and holistic solutions in this tension field, the orchestration of different subservices is necessary, bundling the functionality of individual smart devices and models. For this goal to be achieved, a framework that considers the complex challenges of the ecosystem focusing on the adoption and operation of smart services is required. Here our paper makes a key contribution. Based on requirements from the literature and concrete smart living use cases, we derive a platform framework for this data ecosystem.
  • Konferenzbeitrag
    Understanding and addressing user needs for annotation of simple sensor data: Bridging the gap between human sensemaking and machine interpretation
    (INFORMATIK 2024, 2024) Kurze, Albrecht; Reuter, Christin
    The increasing presence of sensors in smart homes generates vast amounts of data, which require effective interpretation to be useful, often along with data annotation. While automatic approaches can automatically analyze sensor data but require strict and clean annotations, they often neglect the complex, multidimensional nature of human sensemaking. We explore this gap and propose an approach to bridge this gap. We present preliminary findings from three directions: lay user annotations of sensor data collected in a field study using our Sensorkit solution, analysis of existing annotation tools, and a human-centered design process for a new annotation solution. Our goal is to develop a more integrated approach to sensor data interpretation that benefits both humans and machines.
  • Konferenzbeitrag
    Understanding stegomalware in ICS: Attacks and Prevention
    (INFORMATIK 2024, 2024) Edeh, Natasha; Yatagha, Romarick; Mejri, Oumayma; Waedt, Karl
    This research investigates the growing threat of stego-malware in Industrial Control Systems (ICS), where attackers utilize steganography to embed malicious code covertly. Such attacks pose significant challenges due to their ability to evade traditional detection methods. The study reviews current cybersecurity frameworks and detection techniques, highlighting their strengths and limitations against stego-malware. It explores various detection approaches, including signature-based, anomaly-based, and AI/ML-based methods, assessing their effectiveness within the context of ISO/IEC 27001 and IEC 62443 standards. Case studies such as Havex and Industroyer underscore the real-world impact of stego-malware on ICS infrastructure. The research advocates for enhanced integration of AI and machine learning to bolster steganalysis capabilities, and proposes improvements to existing cybersecurity frameworks to address steganographic threats more effectively. By bridging gaps in current knowledge, this study contributes to advancing cybersecurity measures tailored to protect critical ICS environments against evolving cyber threats.