TY - THES A1 - Quinzan, Francesco T1 - Combinatorial problems and scalability in artificial intelligence N2 - Modern datasets often exhibit diverse, feature-rich, unstructured data, and they are massive in size. This is the case of social networks, human genome, and e-commerce databases. As Artificial Intelligence (AI) systems are increasingly used to detect pattern in data and predict future outcome, there are growing concerns on their ability to process large amounts of data. Motivated by these concerns, we study the problem of designing AI systems that are scalable to very large and heterogeneous data-sets. Many AI systems require to solve combinatorial optimization problems in their course of action. These optimization problems are typically NP-hard, and they may exhibit additional side constraints. However, the underlying objective functions often exhibit additional properties. These properties can be exploited to design suitable optimization algorithms. One of these properties is the well-studied notion of submodularity, which captures diminishing returns. Submodularity is often found in real-world applications. Furthermore, many relevant applications exhibit generalizations of this property. In this thesis, we propose new scalable optimization algorithms for combinatorial problems with diminishing returns. Specifically, we focus on three problems, the Maximum Entropy Sampling problem, Video Summarization, and Feature Selection. For each problem, we propose new algorithms that work at scale. These algorithms are based on a variety of techniques, such as forward step-wise selection and adaptive sampling. Our proposed algorithms yield strong approximation guarantees, and the perform well experimentally. We first study the Maximum Entropy Sampling problem. This problem consists of selecting a subset of random variables from a larger set, that maximize the entropy. By using diminishing return properties, we develop a simple forward step-wise selection optimization algorithm for this problem. Then, we study the problem of selecting a subset of frames, that represent a given video. Again, this problem corresponds to a submodular maximization problem. We provide a new adaptive sampling algorithm for this problem, suitable to handle the complex side constraints imposed by the application. We conclude by studying Feature Selection. In this case, the underlying objective functions generalize the notion of submodularity. We provide a new adaptive sequencing algorithm for this problem, based on the Orthogonal Matching Pursuit paradigm. Overall, we study practically relevant combinatorial problems, and we propose new algorithms to solve them. We demonstrate that these algorithms are suitable to handle massive datasets. However, our analysis is not problem-specific, and our results can be applied to other domains, if diminishing return properties hold. We hope that the flexibility of our framework inspires further research into scalability in AI. N2 - Moderne Datensätze bestehen oft aus vielfältigen, funktionsreichen und unstrukturierten Daten, die zudem sehr groß sind. Dies gilt insbesondere für soziale Netzwerke, das menschliche Genom und E-Commerce Datenbanken. Mit dem zunehmenden Einsatz von künstlicher Intelligenz (KI) um Muster in den Daten zu erkennen und künftige Ergebnisse vorherzusagen, steigen auch die Bedenken hinsichtlich ihrer Fähigkeit große Datenmengen zu verarbeiten. Aus diesem Grund untersuchen wir das Problem der Entwicklung von KI-Systemen, die auf große und heterogene Datensätze skalieren. Viele KI-Systeme müssen während ihres Einsatzes kombinatorische Optimierungsprobleme lösen. Diese Optimierungsprobleme sind in der Regel NP-schwer und können zusätzliche Nebeneinschränkungen aufwiesen. Die Zielfunktionen dieser Probleme weisen jedoch oft zusätzliche Eigenschaften auf. Diese Eigenschaften können genutzt werden, um geeignete Optimierungsalgorithmen zu entwickeln. Eine dieser Eigenschaften ist das wohluntersuchte Konzept der Submodularität, das das Konzept des abnehmende Erträge beschreibt. Submodularität findet sich in vielen realen Anwendungen. Darüber hinaus weisen viele relevante An- wendungen Verallgemeinerungen dieser Eigenschaft auf. In dieser Arbeit schlagen wir neue skalierbare Algorithmen für kombinatorische Probleme mit abnehmenden Erträgen vor. Wir konzentrieren uns hierbei insbesondere auf drei Prob- leme: dem Maximum-Entropie-Stichproben Problem, der Videozusammenfassung und der Feature Selection. Für jedes dieser Probleme schlagen wir neue Algorithmen vor, die gut skalieren. Diese Algorithmen basieren auf verschiedenen Techniken wie der schrittweisen Vorwärtsauswahl und dem adaptiven sampling. Die von uns vorgeschlagenen Algorithmen bieten sehr gute Annäherungsgarantien und zeigen auch experimentell gute Leistung. Zunächst untersuchen wir das Maximum-Entropy-Sampling Problem. Dieses Problem besteht darin, zufällige Variablen aus einer größeren Menge auszuwählen, welche die Entropie maximieren. Durch die Verwendung der Eigenschaften des abnehmenden Ertrags entwickeln wir einen einfachen forward step-wise selection Optimierungsalgorithmus für dieses Problem. Anschließend untersuchen wir das Problem der Auswahl einer Teilmenge von Bildern, die ein bestimmtes Video repräsentieren. Dieses Problem entspricht einem submodularen Maximierungsproblem. Hierfür stellen wir einen neuen adaptiven Sampling-Algorithmus für dieses Problem zur Verfügung, das auch komplexe Nebenbedingungen erfüllen kann, welche durch die Anwendung entstehen. Abschließend untersuchen wir die Feature Selection. In diesem Fall verallgemeinern die zugrundeliegenden Zielfunktionen die Idee der submodularität. Wir stellen einen neuen adaptiven Sequenzierungsalgorithmus für dieses Problem vor, der auf dem Orthogonal Matching Pursuit Paradigma basiert. Insgesamt untersuchen wir praktisch relevante kombinatorische Probleme und schlagen neue Algorithmen vor, um diese zu lösen. Wir zeigen, dass diese Algorithmen für die Verarbeitung großer Datensätze geeignet sind. Unsere Auswertung ist jedoch nicht problemspezifisch und unsere Ergebnisse lassen sich auch auf andere Bereiche anwenden, sofern die Eigenschaften des abnehmenden Ertrags gelten. Wir hoffen, dass die Flexibilität unseres Frameworks die weitere Forschung im Bereich der Skalierbarkeit im Bereich KI anregt. KW - artificial intelligence KW - scalability KW - optimization KW - Künstliche Intelligenz KW - Optimierung Y1 - 2023 U6 - http://nbn-resolving.de/urn/resolver.pl?urn:nbn:de:kobv:517-opus4-611114 ER - TY - JOUR A1 - Siddiqi, Muhammad Ali A1 - Dörr, Christian A1 - Strydis, Christos T1 - IMDfence BT - architecting a secure protocol for implantable medical devices JF - IEEE access N2 - Over the past decade, focus on the security and privacy aspects of implantable medical devices (IMDs) has intensified, driven by the multitude of cybersecurity vulnerabilities found in various existing devices. However, due to their strict computational, energy and physical constraints, conventional security protocols are not directly applicable to IMDs. Custom-tailored schemes have been proposed instead which, however, fail to cover the full spectrum of security features that modern IMDs and their ecosystems so critically require. In this paper we propose IMDfence, a security protocol for IMD ecosystems that provides a comprehensive yet practical security portfolio, which includes availability, non-repudiation, access control, entity authentication, remote monitoring and system scalability. The protocol also allows emergency access that results in the graceful degradation of offered services without compromising security and patient safety. The performance of the security protocol as well as its feasibility and impact on modern IMDs are extensively analyzed and evaluated. We find that IMDfence achieves the above security requirements at a mere less than 7% increase in total IMD energy consumption, and less than 14 ms and 9 kB increase in system delay and memory footprint, respectively. KW - protocols KW - implants KW - authentication KW - ecosystems KW - remote monitoring KW - scalability KW - authentication protocol KW - battery-depletion attack KW - battery KW - DoS KW - denial-of-service attack KW - IMD KW - implantable medical device KW - non-repudiation KW - smart card KW - zero-power defense Y1 - 2020 U6 - https://doi.org/10.1109/ACCESS.2020.3015686 SN - 2169-3536 VL - 8 SP - 147948 EP - 147964 PB - Institute of Electrical and Electronics Engineers CY - Piscataway ER - TY - JOUR A1 - Ghahremani, Sona A1 - Giese, Holger A1 - Vogel, Thomas T1 - Improving scalability and reward of utility-driven self-healing for large dynamic architectures JF - ACM transactions on autonomous and adaptive systems N2 - Self-adaptation can be realized in various ways. Rule-based approaches prescribe the adaptation to be executed if the system or environment satisfies certain conditions. They result in scalable solutions but often with merely satisfying adaptation decisions. In contrast, utility-driven approaches determine optimal decisions by using an often costly optimization, which typically does not scale for large problems. We propose a rule-based and utility-driven adaptation scheme that achieves the benefits of both directions such that the adaptation decisions are optimal, whereas the computation scales by avoiding an expensive optimization. We use this adaptation scheme for architecture-based self-healing of large software systems. For this purpose, we define the utility for large dynamic architectures of such systems based on patterns that define issues the self-healing must address. Moreover, we use pattern-based adaptation rules to resolve these issues. Using a pattern-based scheme to define the utility and adaptation rules allows us to compute the impact of each rule application on the overall utility and to realize an incremental and efficient utility-driven self-healing. In addition to formally analyzing the computational effort and optimality of the proposed scheme, we thoroughly demonstrate its scalability and optimality in terms of reward in comparative experiments with a static rule-based approach as a baseline and a utility-driven approach using a constraint solver. These experiments are based on different failure profiles derived from real-world failure logs. We also investigate the impact of different failure profile characteristics on the scalability and reward to evaluate the robustness of the different approaches. KW - self-healing KW - adaptation rules KW - architecture-based adaptation KW - utility KW - reward KW - scalability KW - performance KW - failure profile model Y1 - 2020 U6 - https://doi.org/10.1145/3380965 SN - 1556-4665 SN - 1556-4703 VL - 14 IS - 3 PB - Association for Computing Machinery CY - New York ER -