University of Thessaly (UTH)
Data and uncertainty driven optimization for dymamic edge computing
Abstract
dc:descriptionIn the era of digital transformation, the convergence of ground breaking computing paradigms, namely the Internet of Things, EdgeComputing, Cluster Computing, and Pervasive Computing stands at the forefront of technological evolution. These paradigms are not only reshaping the technological landscape, but are also redefining the boundaries of efficiency, scalability, resource optimization, and adaptability in distributed systems. This thesis embarks on an in-depth exploration of these interconnected technologies, focusing on enhancing data processing and machine learning efficiency in edge computing environments, driven by both data characteristics and uncertainty. The research develops novel frameworks and algorithms in five key areas: (1) a synopsis-based similarity extraction framework that enables distributed nodes to identify and collaborate with peers holding similar datasets, optimizing data exchange and reducing redundancy in distributed environments; (2) a data-aware training acceleration framework that optimizes machine learning model training in resource-constrained environments, reducing the training time and improving efficiency while maintaining models’ performance over time; (3) a transfer learning model that enhances knowledge transfer between nodes in edge environments, leveraging model selection techniques to improve accuracy and efficiency in resource-constrained settings; (4) a drift-aware task management mechanism that dynamically adapts task allocation strategies based on evolving data distributions, ensuring efficient resource utilization and system adaptability in edge computing; (5) a correlation-aware task scheduling approach that optimally assigns tasks by detecting dependencies and resource availability, improving execution efficiency and minimizing bottlenecks in distributed systems. These contributions provide robust solutions to critical challengesin edge and cluster computing environments, where traditional methods struggle with scalability, resource constraints, and real-time data evolution. By integrating advanced statistical methods, adaptive learning, and task optimization techniques, the frameworks proposed in this thesis pave the way for more efficient and adaptive distributed systems. These developments contribute to the ongoing evolution of Internet of Things, Edge Computing, and related domains, offering practical applications in smart cities, autonomous systems, and large scale industrial networks.
Degree
thesis:*- Grantor dc:publisher
- University of Thessaly (UTH)
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Moustakas, Athanasios
- Μουστάκας, Αθανάσιος
Subjects
dc:subject × 26- Edge computing
- Internet of Things
- Cluster computing
- Machine learning
- Distributed systems
- Data-driven optimization
- Uncertainty-driven optimization
- Υπολογισμός στο Άκρο
- Διαδίκτυο των Πραγμάτων
- Υπολογιστικά συστήματα συστάδων
- Μηχανική μάθηση
- Κατανεμημένα συστήματα
- Βελτιστοποίηση με βάση τα δεδομένα
- Βελτιστοποίηση με βάση την αβεβαιότητα
- Μαθηματικά
- Φυσικές Επιστήμες
- Έλεγχος και Βελτιστοποίηση
- Στατιστική και Πιθανότητες
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Τεχνητή νοημοσύνη
- Mathematics
- Natural Sciences
- Control and Optimization
- Statistics and Probability
- Computer and Information Sciences
- Artificial Intelligence
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/59858