P314 - INFORMATIK 2021 - Computer Science & Sustainability
Autor*innen mit den meisten Dokumenten
Auflistung nach:
Neueste Veröffentlichungen
- TextdokumentFrom Physical to Virtual: Leveraging Drone Imagery to Automate Photovoltaic System Maintenance(INFORMATIK 2021, 2021) Lowin, Maximilian; Kellner, Domenic; Kohl, Tobias; Mihale-Wilson, CristinaOptimizing the maintenance of large-scale infrastructure can be a significant cost driver for small and medium-sized enterprises (SMEs). This paper presents a feasible approach to combine data from real-world physical structures collected through an automated maintenance process with cloud-based AI services to generate a meaningful virtual representation of such structures. We use photovoltaic systems as an exemplary physical structure and thermal imaging, collected through scheduled drone monitoring. With help of these unstructured data sources, we demonstrate our approach's applicability. Our solution artifact provides a lightweight AI application that is adoptable for other problem spaces, enabling an easier knowledge transfer from research to SMEs. By combining Cloud Computing with Machine Learning, the artifact identifies present and emerging damages of physical objects. It provides a virtual representation of the object's state and empowers a meaningful visualization.
- TextdokumentEnergieeffizientes Kaltstartverhalten spanender Werkzeugmaschinen(INFORMATIK 2021, 2021) Walz, Deborah; Wächter, Andreas; Tomov, Stefan; Heimbach, Konrad; Weigold, MatthiasDie Kompensation thermischer Einflüsse und daraus resultierender geometrischer Verlagerungen spielt eine bedeutende Rolle bei der Gewährleistung einer hohen Bearbeitungsqualität von Werkstücken in Zerspanungsprozessen. Übliche Vorgehensweisen zur Reduktion thermischer Verlagerungen während der Produktion gehen mit einem erheblichen Energiebedarf einher oder modellieren die komplexen Zusammenhänge thermischer Einflüsse nur ungenügend. Methoden des Maschinellen Lernens stellen einen vielversprechenden Ansatz zur Modellierung dar. Es wird eine Lösung angestrebt, die aufwandsarm auf Produktionsmaschinen ähnlicher Bauart übertragen werden kann. Derzeit ist ungeklärt, ob eine explizite oder implizite Modellierung der zeitlich multivarianten Daten eine ufriedenstellende Lösung bietet. Als besonders herausfordernd stellt sich die Verfügbarkeit von ausreichend vielen Datenbeispielen zur Modellierung der relevanten Größen dar.
- TextdokumentChameleon: A Semi-AutoML framework targeting quick and scalable development and deployment of production-ready ML systems for SMEs(INFORMATIK 2021, 2021) Otterbach, Johannes; Wollmann, ThomasDeveloping, scaling, and deploying modern Machine Learning solutions remains challenging for small- and middle-sized enterprises (SMEs). This is due to a high entry barrier of building and maintaining a dedicated IT team as well as the difficulties of real-world data (RWD) compared to standard benchmark data. To address this challenge, we discuss the implementation and concepts of Chameleon, a semi-AutoML framework. The goal of Chameleon is fast and scalable development and deployment of production-ready machine learning systems into the workflow of SMEs. We first discuss the RWD challenges faced by SMEs. After, we outline the central part of the framework which is a model and loss-function zoo with RWD-relevant defaults. Subsequently, we present how one can use a templatable framework in order to automate the experiment iteration cycle, as well as close the gap between development and deployment. Finally, we touch on our testing framework component allowing us to investigate common model failure modes and support best practices of model deployment governance.
- TextdokumentBrezel-Cast: Verkaufsprognose von Backwaren(INFORMATIK 2021, 2021) Döring, Nico; Kreiss, Jonathan; Schuster, Thomas; Volz, RaphaelIn diesem Papier diskutieren wir die Anwendbarkeit von Verfahren der künstlichen Intelligenz zur Prognose von Absatzzahlen für eine Bäckerei mittlerer Größe. Dabei wird beschrieben, wie bei der Entwicklung zusätzliche Daten (Kontextinformationen) zur Prognose genutzt werden. Daraufhin werden zwei Verfahren des maschinellen Lernens trainiert und im Ergebnis miteinander verglichen. Neben einer abschließenden Bewertung und Ausblick auf zukünftige Verbesserungen, wird zudem eine Einschätzung zum Einsatz im Produktivbetrieb abgegeben.
- TextdokumentBlueprint for a Production-Ready Information Retrieval System based on Multi-Modal Embeddings(INFORMATIK 2021, 2021) Ebert, André; Apel, Anika; Chodyko, Piotr; Hiroyasu, Kyle; Ismali, Festina; Koo, Hyein; Kronburger, Julia; Pesch, RobertDeep Learning models for mapping documents from different domains, e.g., text, images, and audio, into a common vector space, enable a seamless information retrieval between the different domains and, thus, significantly improve the user experience of many expert tools. Despite various models for multi-modal mappings presented in scientific literature, the implementation and integration remain a challenge within the industry, especially for small or medium-sized companies. Reasons are, that developing such retrieval systems for production use-cases is a non-trivial task, requiring scalable, reliable, and cost-efficient infrastructure, services as well as adequate Deep Learning models. We present a generic and flexible blueprint architecture, targeting the development of a production-ready image-text retrieval search system using Kubernetes, MLflow, Elasticsearch, and integrating state-of-the-art Deep Learning models.
- TextdokumentKontinuierliche Evaluierung eines KI-basierten interaktiven Systems(INFORMATIK 2021, 2021) Mummert, Niclas; Lange, Olga; Cioflica, Paul; Reutemann, TobiasEine kontinuierliche Evaluierung eines KI-basierten interaktiven Systems wird in diesem Beitrag am Beispiel eines Fahrerassistenzsystems zur Ermittlung von Nutzerpräferenzen über eine Auswahl vom Fahrmodus (Comfort, Normal Sport und Sport+) aus dem Automobilbau vorgestellt. Das zu evaluierende System enthält Machine Learning Modelle, welche anhand Straßen (Asphalt, Pflastersteine, Feldweg) und Nutzerinformationen einen Fahrmodus vorschlägt. Herausforderungen der Evaluierung entstehen vor allem durch die Eingaben der Nutzenden an der Benutzungsschnittstelle (User Interface). Dieser Beitrag stellt mögliche Evaluierungsmethoden zur kontinuierlichen Erfassung der Qualität dieses KI-basierten interaktiven Systems dar und geht der Frage nach Zusammenhängen zwischen dem Handeln der Nutzenden und der Veränderung der ML-Modellen nach.
- TextdokumentKünstliche Intelligenz und Nachhaltigkeit(INFORMATIK 2021, 2021) Mainzer, KlausDie Informatik hat sich in den letzten Jahrzehnten zu einem der größten Treiber des gesellschaftlichen Wandels entwickelt. Bedingt durch den rasanten Fortschritt in der Informationstechnik finden Informatiksysteme eine rasante Verbreitung. Die Infrastrukturen des Internets und World Wide Webs schaffen völlig neue Formen der Interaktion und Kommunikation zwischen Menschen, Maschinen und Infrastrukturen (Internet der Dinge). Produkte und Dienstleistungen auf Basis von Softwaresystemen greifen um sich. Informatik verändert in vielen Anwendungsgebieten die Sicht auf die Welt. Damit steht die Informatik als Treiber des Wandels auf den ersten Blick in einem krassen Gegensatz zu einer statisch verstandenen Nachhaltigkeit.
- TextdokumentImproving a Rule-based Fraud Detection System with Classification Based on Association Rule Mining(INFORMATIK 2021, 2021) Baumann, MichaelaImproving a Rule-based Fraud Detection System with Classification Based on Association Rule MiningThe detection of fraudulent insurance claims is a great challenge for insurance companies. Although the detection possibilities are getting better and better, fraudsters do not hesitate also using newer and more sophisticated methods. Apart from establishing new fraud detection systems, also the existing systems need to be updated and improved as best as possible. One common detection system is a rule-based expert system that checks predefined rules and gives alerts when certain conditions are met. Usually, the rules are treated separately and correlations within the rules are considered insufficiently. The work at hand describes how the classification based on association rule mining is used for improving such rule-based systems by bringing in relations between pairs of rules. The rule weights are determined through a genetic optimizer.
- TextdokumentOrganizing for temporality and supporting AI systems – a framework for applied AI and organization research(INFORMATIK 2021, 2021) Nuhn, HelgeBoth temporary forms of organizing (TO) and artificial intelligence have recently received increased practical and scholarly attention [HW19],[Zh19]. Their combination has not yet been subject of research or research application, nor is there a roadmap to the development of TOspecific AI applications. In relation to permanent organizations, TOs devote more time to organizing, planning, and adapting to change, but supporting organizing as a task is not yet in the scope of AI research. This article creates new links between the domains computer science and organizational theory. It reviews TOs as special vehicles for organizing endeavors, proposes relevant properties, reviews recent AI research advances, and synthesizes challenges for researching into AIassisted organizing in TOs. A table of use cases along with proposed AI methods and required data proposes future research activities and is basis for a call for data sets, data challenges and metrics for assessing AI-assisted organizing, especially in temporary contexts.
- TextdokumentCan smartphone generated cycling data contribute to the improvement of the bicycle infrastructure?(INFORMATIK 2021, 2021) Höper, Tobias Höper; Schering, Johannes; Marx Gómez, JorgeSmartphone generated data could make an important contribution to learn more about cycling behaviors to support the decision making process. As part of the Oldenburg Bicycle Challenge a huge amount of app based data was gathered in the German Oldenburg region. In combination with weather and counting data new knowledge on bicycle use can be generated. Parameters as distances and durations of the trips, hourly values, the distribution of the trips to recreational and utilitarian purposes, brakings and waiting times are analyzed and discussed as part of this work. GPS data of cyclists is not only gathered by smartphone apps but also by bike sensors that are attached to the bicycle as many projects has shown (e.g. ECOSense). As part of the conclusion, the advantages and disadvantages of both approaches regarding acquisition of participants, effort for data collection, data quality, potentials for cycling data analysis etc. will be discussed.