14043 research outputs found
Sort by
An analysis of the microstructural and functional network architecture of human intelligence
No full textIntelligenz ist mit zahlreichen Hirnregionen und Tausenden genetischen Varianten assoziiert. Dennoch sind Mechanismen der Intelligenzunterschiede unklar. Diese Dissertation untersuchte die mikrostrukturelle und funktionelle Netzwerkarchitektur der Intelligenz genauer. Studie eins deckte mittels Multicenter-Design robuste positive Assoziationen zwischen fraktioneller Anisotropie und Intelligenz in drei Clustern auf. Studie zwei untersuchte, ob die Neuritendichte, die Orientierungsdispersion oder der Myelinwasseranteil von Fasertrakten der weißen Substanz Effekte polygener Scores auf Intelligenz vermittelten und identifizierte die Neuritendichte spezifischer Fasertrakte als mediierende Eigenschaft. Studien drei und vier analysierten die funktionelle Netzwerkkonnektivität im Ruhezustand. Während Studie drei keine robusten Zusammenhänge zwischen Metriken basierend auf fMRT-Daten und Intelligenz ergab, zeigte Studie vier frequenzspezifische Mediationseffekte basierend auf EEG-Daten
Predictable feature analysis and slow feature machines
No full textUm interaktive Aufgaben mit Slow Feature Analysis (SFA) und ähnlichen Methoden zu behandeln, wird Predictable Feature Analysis (PFA) entwickelt - ein Algorithmus, der möglichst vorhersagbare Komponenten aus hochdimensionalen Eingabedaten extrahiert. Mit Hilfe dieser kann ein System lokal gesteuert werden und durch Einführung von "Slow Feature Machines" (SFMs) lässt sich sogar globale Steuerung erreichen: SFA liefert eine monotone Darstellung, die mittels lokaler Methoden global optimiert werden kann. Dies wird experimentell getestet und weiterentwickelt, z.B.: Erweiterung um ein Gedächtnis; Einsatz von SFA zur Datensynthese und Steuerung; Verbesserung von xSFA durch Entwicklung von "fast xSFA" (fxSFA) und weiteren Methoden.
Zur mathematischen Analyse wird die existierende SFA-Theorie in riemannscher Geometrie formuliert und eine Verbindung zum Neumann-Laplace-Operator hergestellt. SFA-Theorie wird auch auf periodische Datenquellen, auf den singulären Fall und anderweitig erweitert
Daten-getriebene Methoden zur Analyse von Alarmfluten in automatisierten industriellen Prozessen
No full textDie Prozessindustrie steht vor der Herausforderung, komplexe Prozesse sicher und effizient zu betreiben. Alarmsysteme unterstützen hierbei, doch sogenannte Alarmfluten – durch eine Vielzahl konsekutiver Alarme – können das Situationsbewusstsein beeinträchtigen. Diese Dissertation entwickelt ein modulares Framework zur datengetriebenen Analyse von Alarmfluten, das Erkennung, Clustering, Klassifikation, Ursachenanalyse und Unsicherheitsbewertung integriert. Zentrale Beiträge sind robuste Detektionsverfahren, fortschrittliche Merkmalsextraktion, erklärbare KI zur Entscheidungsunterstützung sowie dynamische Kausalanalysen. Zudem werden drei öffentlich zugängliche Datensätze (Tennessee-Eastman-Prozess, Kernkraftwerk, synthetische Szenarien) vorgestellt, um reproduzierbare Evaluierungen zu ermöglichen. Das Framework überwindet Einschränkungen bestehender Ansätze, verbessert das Verständnis von Alarmdynamiken und Abhängigkeiten und trägt so zu höherer Betriebssicherheit und Effizienz bei.The process industry faces the challenge of operating complex processes safely and efficiently. Alarm systems provide essential support, yet so-called alarm floods—triggered by large numbers of consecutive alarms—can impair operators’ situational awareness. This dissertation presents a modular framework for data-driven alarm flood analysis, integrating detection, clustering, classification, root cause analysis, and uncertainty assessment. Key contributions include robust detection methods, advanced feature extraction, explainable AI for decision support, and dynamic causal analysis. Furthermore, three publicly available datasets (Tennessee Eastman process, nuclear power plant, synthetic scenarios) are introduced to enable reproducible evaluation. The framework addresses limitations of existing approaches, enhances the understanding of alarm dynamics and dependencies, and contributes to improved operational safety and efficiency
Relevance, breadth, or depth?
No full textProductive Failure (PF) und Vicarious Failure (VF) sind Vorbereitungsaktivitäten im Rahmen von Problemlösen vor Instruktion. Durch Generieren (PF) oder Analysieren (VF) fehlerhafter Lösungsansätze zu einem unbekannten Problem soll konzeptuelles Wissen aufgebaut werden. Ein zentraler, aber bislang unzureichend untersuchter Vorbereitungsmechanismus ist die Aktivierung von Vorwissen. Diese Dissertation untersuchte die Rolle von Breite und Relevanz der Vorwissensaktivierung in PF und VF in drei Studien. Studie 1 zeigte weitgehende Vergleichbarkeit beider Aktivitäten in Lernprozessen. Studie 2 manipulierte die Breite relevanter Vorwissensaktivierung mittels Lösungsansätzen in VF, Studie 3 die Breite und Relevanz mittels Zielformulierungen in PF. Nicht Breite, sondern Relevanz erwies sich als entscheidend für lernförderliche Vorwissensaktivierung. Zudem fungierte intermediäres Wissen als Bindeglied zwischen Aktivierung und Instruktion. are preparatory activities of the instructional design problem solving prior to instruction. By generating (PF) or analyzing (VF) erroneous solution attempts to an unfamiliar problem, students build conceptual knowledge. A central, yet insufficiently examined, preparatory mechanism is the activation of prior knowledge during the preparatory activity. This dissertation investigated the role of breadth and relevance of prior knowledge activation in PF and VF in three studies. Study 1 demonstrated comparability of the two activities in their learning processes. Study 2 manipulated the breadth of relevant prior knowledge activation via solution approaches in VF, while Study 3 manipulated both breadth and relevance through goal formulations in PF. Not breadth, but relevance was central for learning. Furthermore, intermediate knowledge functioned as a link between prior knowledge activation and learning from instruction
Experimental validation of the effective stress concept in unsaturated soils using the suction stress approach across multiple test types and a wide suction range
No full textIm ungesättigten Zustand befinden sich Wasser und Luft im Porenraum des Bodens. Dieses führt zu zusätzlichen Kräften im Korngefüge, die die Festigkeit des Bodens erhöhen. Im Konzept der effektiven Spannungen kann dieses durch die Spannungsvariable "Suction Stress" erfasst werden. Wenn das effektive Spannungskonzept für ungesättigten Böden anwendbar ist, sollten Werte für den "Suction Stress" abgeleitet aus verschiedenen Versuchsarten konsistent sein. Es wurden Biaxial-, direkte Scher-, einaxiale Druck- und Zugversuche an einem Boden über einen großen Bereich von Saugspannungen durchgeführt und jeweils der "Suction Stress" bestimmt. Erkenntnisse sind der signifikante Anstieg des Reibungswinkels bei hohen Saugspannungen, der im Vergleich zu dem "Suction Stress" aus Scherversuchen unter Druckbelastung deutlich niedrigere "Suction Stress" aus Zugversuchen, abweichende Werte des "Suction Stress" ermittelt aus Biaxial-, direkten Scher- und einaxialen Druckversuchen bei hohen Saugspannungen
Worst-case time analysis of key agreement protocols in 10BASE-T1S automotive networks
No full textWith the rise of in-vehicle and car-to-x communication systems, ensuring robust security in automotive networks is becoming increasingly vital. As the industry shifts toward Ethernet-based architectures, the IEEE 802.1AE MACsec standard is gaining prominence as a critical security solution for future in-vehicle networks (IVNs). MACsec utilizes the MACsec Key Agreement Protocol (MKA), defined in the IEEE 802.1X standard, to establish secure encryption keys for data transmission. However, when applied to 10BASE-T1S Ethernet networks with multidrop topologies, MKA encounters a significant challenge known as the real-time paradox. This paradox arises from the competing demands of prioritizing key agreement messages and real-time control data, which conflict with each other. Infineon addresses this challenge with its innovative In-Line Key Agreement (IKA) protocol. By embedding key agreement information directly within a standard data frame, IKA effectively resolves the real-time paradox and enhances network performance. This paper establishes a theoretical worst-case delay bound for key agreement in multidrop 10BASE-T1S IVNs with more than two nodes, using Network Calculus techniques. The analysis compares the MKA and IKA protocols in terms of performance. For a startup scenario involving a 16-node network with a 50 bytes MPDU size, the MKA protocol exhibits a worst-case delay that is 1080 % higher than that of IKA. As the MPDU size increases to 1486 bytes, this performance gap narrows significantly, reducing the delay difference to just 6.6 %
Crypto-agility in automotive real-time systems in context of post-quantum-cryptography
No full textAdvances in quantum computers have brought cryptographic agility into the focus of cyber security. Increasing the key lengths of existing asymmetric algorithms is not enough, as quantum computers can solve the mathematical problems of asymmetric cryptography more efficiently. The algorithms concerned must be replaced in order to guarantee the necessary security. Crypto-agility is particularly relevant for systems that are supplied with software updates over many years, such as modern automobiles based on the concept of Software Defined Vehicles (SDVs). These vehicles define their functionality through software that is continuously expanded. Resource-limited systems in real-time applications that rely on microcontrollers with hardware support for cryptographic algorithms pose a particular challenge. These hardware accelerators cannot be updated, which is a problem in the context of crypto-agility. Therefore, this article presents a concept that enables the update of Electronic Control Units (ECUs) for real-time applications on the AUTomotive Open System ARchitecture (AUTOSAR) Classic Platform. All post-quantum algorithms standardized by National Institute of Standards and Technology (NIST) are implemented and evaluated on two generations of automotive microcontrollers
Static application security testing by abstract interpretation
No full textStatic code analysis has evolved to be a standard technique in the development process of safety-critical software. Its formal method, Abstract Interpretation, supports formal soundness proofs, showing that no defect from the covered defect classes is missed, and scales to real-life industry applications. It can be applied to show compliance to coding guidelines, and to demonstrate the absence of critical vulnerabilities, including "unforgivable defects" like runtime errors and data races. Soundness of the underlying analysis is an essential property since it provides full data and control coverage; in particular it can guarantee that all data and function pointer values have been taken into account. While in the past, sound static analyzers have been primarily applied to demonstrate classical safety properties they are well suited also to discover complex flow-dependent cybersecurity vulnerabilities at the source code level. One important building block is a sound non-interference analysis that can demonstrate the independence between memory locations and can be applied to flexibly model cybersecurity vulnerabilities. This article gives an overview of static analysis by Abstract Interpretation with a focus on taint-based non-interference analysis and its application to detect cybersecurity vulnerabilities caused by data and control dependences. We report on experimental results obtained with the Juliet benchmark suite
BaTiO–SrTiO composites
No full textComposites of ferroelectric and paraelectric perovskites have attracted a lot of attention due to their application potential in energy storage as well as novel computing and memory devices. So far the main focus of research has been on superlattices and ferroelectric particles in a paraelectric matrix, while the impact of paraelectric inclusions on a ferroelectric matrix is surprisingly underrepresented. To close this gap in knowledge we perform molecular dynamics simulations using an derived effective Hamiltonian for BaTiO–SrTiO and reveal the dependency of phase stability and phase transitions on the size and distances of paraelectric inclusions. We discuss how the combination of compressive strain and depolarization fields at the SrTiO interfaces induces large local polarization, complex domain structures and coexisting phases as well as diffuse phase transitions and reduced coercive fields
Risk factors for non-benefit of implantable cardioverter defibrillator therapy
No full textStudies have demonstrated overall prognostic benefits of ICD implantation in patients at increased risk of sudden cardiac death. However, results are inconsistent in certain subgroups. This study aims to evaluate the prognostic implications of comorbidities on ICD outcomes and compare trends in patient selection and outcomes over a decade-long inclusion period. This study analysed 422 patients undergoing ICD implantation between 2011 and 2020. The study endpoint "no-benefit" was characterized by death from any cause occurring without prior appropriate ICD therapy. Benefit of ICD implantation was defined as either receiving appropriate ICD therapy before death or surviving until the end of the observation period. During a mean follow-up of 4.2 3.0 years, no-benefit of ICD implantation was observed in 84 patients (20%). Independent risk factors for no-benefit were age 68 years (HR 4.599, < 0.001), anemia (HR 2.549, < 0.001), peripheral artery disease (HR 2.066, = 0.007), and chronic obstructive pulmonary disease (HR 1.939, = 0.014). Subgroup analysis by age < 68 years and 68 years demonstrated that the risk of no-benefit increases with age and comorbidities. When comparing patients with ICD implantation in 2011–2015 with those in 2016–2020, there were no significant differences in one-, two- and three-year-no-benefit rates. Different comorbidities were associated with no-benefit in the early and late implantation groups. Risk factors such as older age and specific comorbidities are associated with a higher likelihood of no-benefit from ICD implantation. A careful patient selection and consideration of individual risk factors besides advanced age is important