In networked systems engineering, operational datagathered from sensors or logs can be used to build data-drivenfunctions for performance prediction, anomaly detection, andother operational tasks [1]. Future telecom services will share acommon communication and processing infrastructure in orderto achieve cost-efficient and robust operation. A critical issuewill be to ensure service quality, whereby different serviceshave very different requirements. Thanks to recent advances incomputing and networking technologies we are able to collect andprocess measurements from networking and computing devices,in order to predict and forecast certain service qualities, such asvideo streaming or data stores. In this paper we examine thesetechniques, which are based on statistical learning methods. Inparticular we will analyze traces from testbed measurements andbuild predictive models. A detailed description of the testbed,which is localized at KTH, is given in Section II, as well as in[2]. / Inom nätverk och systemteknik samlas operativ data från sensorer eller loggar som sedan kan användas för att bygga datadrivna funktioner för förutsägelser om prestanda och andra operationella uppgifter [1]. Framtidens teletjänster kommer att dela en gemensam kommunikation och bearbetnings infrastruktur i syfte att uppnå kostnadseffektiva och robusta nätverk. Ett kritiskt problem med detta är att kunna garantera en hög servicekvalitet. Detta problem uppstår till stor del som ett resultat av att olika tjänster har olika krav. Tack vare nyliga avanceringar inom beräkning och nätverksteknologi har vi kunnat samla in användningsmätningar från nätverk och olika datorenheter för att kunna förutspå servicekvalitet för exempelvis videostreaming och lagring av data. I detta arbete undersöker vi data med hjälp av statistiska inlärningsmetoder och bygger prediktiva modeller. En mer detaljerat beskrivning av vår testbed, som är lokaliserad på KTH, finns i [2]. / Kandidatexjobb i elektroteknik 2020, KTH, Stockholm

The Internet of Things (IoT) is comprised of numerous devices that are connected through wired or wireless networks, including sensors and actuators. The number of IoT applications has recently increased dramatically, including Smart Homes, Internet of Vehicles (IoV), Internet of Medical Things (IoMT), Smart Cities, and Wearables. IoT Analytics has reported that the number of connected devices is expected to grow 18% to 14.4 billion in 2022 and will be 27 billion by 2025. Security is a critical issue in today's IoT, due to the nature of the architecture, the types of devices, the different methods of communication (mainly wireless), and the volume of data being transmitted over the network. Furthermore, security will become even more important as the number of devices connected to the IoT increases. However, devices can protect themselves and detect threats with the Intrusion Detection System (IDS). IDS typically use one of two approaches: anomaly-based or signature-based. In this thesis, we define the problems and the particular requirements of securing the IoT environments, and we have proposed a Deep Learning (DL) anomaly-based model with optimal features selection to detect the different potential attacks in IoT environments. We then compare the performance results with other works that have been used for similar tasks. We also employ the idea of reinforcement learning to combine the two different IDS approaches (i.e., anomaly-based and signature-based) to enable the model to detect known and unknown IoT attacks and classify the recognized attacked into five classes: Denial of Service (DDoS), Probe, User-to-Root (U2R), Remote-to-Local (R2L), and Normal traffic. We have also shown the effectiveness of two trending machine-learning techniques, Federated and Transfer learning (FL/TL), over using the traditional centralized Machine and Deep Learning (ML/DL) algorithms. Our proposed models improve the model's performance, increase the learning speed, reduce the amount of data that needs to be trained, and reserve user data privacy when compared with the traditional learning approaches. The proposed models are implemented using the three benchmark datasets generated by the Canadian Institute for Cybersecurity (CIC), NSL-KDD, CICIDS2017, and the CSE-CIC-IDS2018. The performance results were evaluated in different metrics, including Accuracy, Detection Rate (DR), False Alarm Rate (FAR), Sensitivity, Specificity, F-measure, and training and fine-tuning times.

Ce mémoire présente différents facteurs qui influencent le développement de la lecture à vue aux claviers de percussion chez les élèves du secondaire. Les facteurs issus du milieu d'apprentissage qui sont essentiels à la formation d'un percussionniste compétent et autonome sont abordés dans la première partie. Ceux-ci forment une base d'habiletés et de connaissances préalables sur lesquelles la lecture à vue peut se développer sans être freinée par différentes lacunes. Des exercices spécifiques sont exposés dans la deuxième partie pour expliquer comment développer séparément les habiletés nécessaires à la lecture à vue, notamment la connaissance de la structure des lames du clavier, l'utilisation des doigtés, la vision périphérique et le sens kinesthésique. La troisième partie est consacrée aux façons de développer la lecture à vue globale en y faisant interagir les habiletés spécifiques qui la composent. Le travail se termine avec un aide-mémoire énonçant les principes, exercices et procédures abordés.

Computational Fluid Dynamics (CFD) and predictive models are presented in this dissertation that illustrates the detailed electrical characteristics, and the current-voltage (i-v) relationship throughout the preheating process of premixed methane-oxygen (CH4-O2) oxyfuel cutting flame subject to electric bias voltages. As such, the equations describing combustion, electrochemical transport for charged species, and potential are solved through a commercially available finite-volume CFD code. The reactions of the methane-oxygen (CH4 – O2) flame were combined with the GRI 3.0 mechanism and a 25-species reduced mechanism, respectively, and additional ionization reactions that generate three chemi-ions, H3O+, HCO+, and e– , to describe the chemistry of ions in flames. The electrical characteristics such as ion migrations and ion distributions are investigated for a range of electric potential, V ∈ [−10V, +10V ]. Since the physical flame is comprised of twelve Bunsen-like conical flame, inclusion of the third dimension imparts the resolution of fluid mechanics and the interaction among the individual cones. As for developing the predictive models, four different supervised machine learning (ML) algorithms, decision tree (DT), random forest (RF), K-nearest neighbors (KNN), and artificial neural network (ANN), were employed to predict the i-v relationship. An experimental dataset of ≈ 10050 was utilized where a 60:20:20 split was adopted, allocating 60% for training, 20% for validation, and 20% for testing. It was concluded that charged 'sheaths' are formed at both torch and workpiece surfaces, subsequently forming three distinct regimes in the i-v relationship. The i-v characteristics obtained have been compared to the previous experimental study for premixed flame. In this way, the overall model generates a better understanding of the physical behavior of the oxyfuel cutting flames, along with a more validated i-v characteristics. Such understanding might provide critical information towards achieving an autonomous oxyfuel cutting process. / Doctor of Philosophy / Oxyfuel flame cutting is a century-old technique having widespread applications in heavy industries, including, but not limited to, building construction, defense, shipyards, etc. However, the mechanized oxyfuel cutting process has never benefited from the degree of autonomy due to contemporary sensing technologies' limitations at high-temperature working conditions. As a result, an experienced labor force is required to operate the system, thereby lowering the efficacy associated with this cutting process. A potential solution to this problem is motivated by preliminary measurements demonstrating that electrical events called 'ion currents' associated with the flame itself can reliably indicate vital process states. Provided that an autonomous process is achieved, this work could realize reliable cost-effective control of the oxyfuel cutting process, a capability of great interest to many core US industries involved in construction, and major equipment manufacture for defense and energy applications. Critical parameters (standoff, F/O ratio, flow rate, etc.) must be detected during operation to ensure an autonomous oxyfuel cutting process. The motivation stems from the fact that by measuring such co-dependence between critical parameters and electrical characteristics through a data acquisition unit (DAQ) and power supply, the shortcomings of sensing suites in a harsh operating environment can be compromised. Experimental data in the literature indicated the current-voltage (i-v) relationship with different critical parameters of oxyfuel flame to be the salient electrical characteristic in the preheating process when cutting steel. A comprehensive two-dimensional computational simulation using StarCCM+ only with the reduced combustion chemical mechanism with ion-exchange reactions has already been completed to elucidate the experimental results and to investigate the electrical characteristics such as ion migrations and ion distributions. Nonetheless, the findings exhibit some magnitude of differences compared to the experimental results. Thereby to further improve the results and better understand the underlying physics, further computational models using ANSYS FLUENT are proposed herein, having the reduced surface chemical mechanism considered. In addition, predictive models were developed based on machine learning (ML) algorithms. Four supervised ML algorithms - decision tree (DT), random forest (RF), Knearest neighbors (KNN), and artificial neural network (ANN) - were adopted to predict the current-voltage (i-v) relationship at different process states. ML offers a more data-driven, adaptable, and scalable approach to prediction compared to traditional methods. Its ability to handle large, noisy, and complex data makes it especially powerful for tasks that are challenging for conventional analytical techniques. The results of this study illustrate the detailed electrical characteristics of premixed methane-oxygen (CH4 – O2) oxyfuel cutting flame subject to an electric field, for both the computational fluid dynamics (CFD) and ML models. Since the physical flame is comprised of twelve Bunsen-like conical flame, inclusion of the third dimension will impart the resolution of fluid mechanics and the interaction among the individual cones. Moreover, the chemical activity at the work surface will also be considered, however, with a substantial simplification of the three-dimensional model as a cost. The overall model will generate a better understanding of the physical behavior of the oxyfuel cutting flames, along with a more validated currentvoltage (i-v) relationship. Consequently, this relationship could then be embedded into a control algorithm to detect the critical process parameters that may facilitate a step towards achieving an autonomous oxyfuel cutting process.

Integrating Terrestrial and Non-Terrestrial Networks (NTNs) within Beyond-5G (B5G) and future 6G ecosystems represents a transformative advancement in achieving ubiquitous, resilient, and scalable communication services. NTNs, including Low Earth Orbit (LEO) satellites, Unmanned Aerial Vehicles (UAVs), and High Altitude Platform Systems (HAPS), extend traditional terrestrial networks by providing continuous connectivity in remote, underserved, and connection-critical scenarios such as disaster-hit regions and rural areas. This thesis deals with an end-to-end cloud-native framework that leverages cutting-edge technologies, including Multi-Access Edge Computing (MEC), Software Defined Networking (SDN), Network Function Virtualization (NFV), blockchain, and advanced AI/ML models, to enhance service availability, security, and Quality of Service (QoS) in 3D NTN environments. The research first explores the deployment of disaggregated Next-Generation Radio Access Networks (NGRANs) across terrestrial and non-terrestrial domains using a Kubernetes-based architecture. A Graph Neural Network (GNN) model is developed to monitor and manage these networks, detecting link failures and optimizing traffic routing paths between terrestrial and satellite components. The GNN model achieves an 85% accuracy in link failure detection and significantly reduces end-to-end delays in NTN deployments, highlighting the potential of AI-driven network management in enhancing overall network resilience. To address the challenge of dynamic resource management in NTNs, this thesis investigates the implementation of functional splits, such as F1 and E1 interfaces, between terrestrial control units (gNB-CU) and satellite-based distributed units (gNB-DU). The study employs Long Short-Term Memory (LSTM) neural networks to predict resource utilization, specifically CPU, memory, and bandwidth of satellite payloads. These predictive models enable proactive monitoring and resource allocation decisions, ensuring efficient use of limited computational resources and maintaining high levels of network performance. Security remains a critical concern in NTNs due to decentralized and open 5G satellite communications. A blockchain-based authentication framework is proposed to mitigate these risks, enhancing the security of data exchanges and remote firmware updates in LEO satellite constellations. Blockchain technology provides a decentralized, transparent, and immutable security framework, improving authentication efficiency and protecting against unauthorized access, though with trade-offs in network performance, such as increased latency and reduced throughput. This approach makes the hybrid B5G NTN network secure, reinforcing the integrity and confidentiality of communication channels, which is essential for emerging services and applications. Furthermore, this thesis comprehensively evaluates MEC-based experimental testbeds that demonstrate service resiliency in NTNs during terrestrial network outages. The MEC deployments allow seamless transitions to satellite access networks, ensuring service continuity and improving QoS. These testbeds showcase the capability of cloud native technologies in maintaining service availability, highlighting their critical role in resilient NTN networks. The findings of this thesis demonstrate that integrating cloud-native architectures, blockchain-based security mechanisms, and advanced AI/ML models significantly enhances the resilience, security, and resource efficiency of NTNs. These innovations pave the way for robust, adaptive, and secure communication systems, supporting the seamless deployment of critical B5G and 6G applications across diverse and challenging environments. This research provides valuable insights into designing and implementing resilient NTNs, setting the foundation for future advancements in global connectivity and intelligent network management.

Cognitive motor integration (CMI), the simultaneous coordination between cerebral function and motor output, is known to deteriorate following a mild traumatic brain injury (mTBI). This thesis explores the relationship between mTBI, CMI, and the performance of elite athletes in the National Hockey League (NHL). The approach focuses on examining the predictive value of various supervised Machine Learning (ML) models with an emphasis on Explainable Artificial Intelligence (XAI) models. Since the ML solution is intended to complement human scouting decisions, we evaluate the experiments based on both interpretability and accuracy on a limited class imbalanced dataset. The contributions of this research are two-fold based on the following research problems: Firstly, the problem of scouting decisions for amateur hockey players to play in the field is addressed by exploring a set of test scores from a neuroscience experiment involving visuomotor performance metrics. Formulated as a supervised binary classification task, results demonstrate that the trained XAI trained models effectively capture the relationship that determines whether amateur hockey players with a history of concussions are likely to play in the NHL. Specifically, we find the best-performing model to be Weighted-Decision Tree trained using all features proposed in this study. Secondly, the effect of previous concussions on scouting decisions is examined by visuomotor metrics and indicators of NHL performance using XAI models. This problem is also formulated as a supervised binary classification task and results show that the trained XAI models are able to predict concussion history using the visuomotor metrics. While results for this question are inconclusive, we give evidence from current neuroscience literature to support why these models do not reach satisfactory performance. Unlike previous research that mainly relies on physical metrics, our work is novel as it utilizes data derived from a neuroscience test, capturing persistent neurocognitive deficits in elite hockey athletes following concussions.

<p dir="ltr">This study introduces a methodology wherein daily values are linearly interpolated to achieve uniform temporal resolution across various data sets, including spectral and environmental information. This approach facilitates further analysis using machine learning techniques to estimate biomass. The proposed Best Friend Frame (B.F.F.) data set integrates Unmanned Aerial Systems (UAS) data, weather data, weather indices, soil hydrological group classifications, and topographic information. Two different biomass estimations were created to enhance versatility: one averaged per management practice and another averaged per physical experimental plot size. Additionally, SuperDove satellite data were combined with identical environmental data as that of the UAS.</p><p dir="ltr">UAS flights were conducted at the ACRE field in 2022 and 2023. The UAS data were captured at a height of 30 meters, yielding a ground sample distance of 2 cm/pixel per flight. Satellite data were sourced from the Planet SuperDove product. The images were processed using Crop Image Extraction (CIE) and calibrated with Vegetation Index Derivation (VID). Spatial resolution was defined as the experimental plot size per year per crop type (soybean or corn). Topographic data were derived from Indiana LiDAR data, and soil information was obtained from the USDA SSURGO dataset.</p><p dir="ltr">The B.F.F. framework can be utilized with various models to identify which environmental inputs influence biomass accumulation throughout the growing season.</p>

Software refactoring is a crucial approach in both development and maintenance to improve the efficiency, maintainability, and structure of software systems. However, a number of challenges remain in the way of the effective implementation of software refactoring techniques within Sri Lanka's software development industry. This thesis investigates the challenger in software refactoring process in Sri Lanka software development companies and examine the intentions of developers, software test automation engineer and project managers on the usage on the machine learning techniques for software refactoring and the study uses the Unified Theory of Acceptance and usage of Technology 2 (UTAUT2) extended model. The study demonstrates that professional in software development Industry have positive intentions toward the usage of machine learning techniques, motivated by benefits they perceive, such as increased productivity, maintenance, and improved code quality. This study advances our understanding of software refactoring and theadoption of new ML technologies and offers insightful information to researchers, practitioners, and decision- makers in the Sri Lankan IT sector and beyond.

Lately, models utilizing channel impulse response (CIR) data for training deep neural networks used for positioning in radio networks have shown promise, particularly in simulated indoor environments. Research has extended to real outdoor setups as well. In this study, deep neural networks originally designed for image classification were employed to estimate positions using both real and simulated outdoor CIR data. A ray tracing simulator was utilized to generate a simulated dataset which corresponded to a real-world dataset. Models were trained and tested on both datasets. To facilitate training on simulated data and testing on real data, a generative adversarial network (GAN) was employed. The thesis concludes that deep neural networks can effectively be used in real outdoor scenarios, but dense data sampling is likely necessary to achieve satisfactory performance across an area. Additionally, it was found that the simulated data used in this study differed significantly from reality, and the employed GAN could not effectively bridge this gap. Consequently, models trained on simulated data performed poorly when tested on real data. However, it was found that deep neural networks significantly outperformed the baseline K-nearest neighbor algorithm when trained and tested on simulated data. However, this was the only case where such a significant advantage for the deep models was observed.

Contrairement à l’idée reçue selon laquelle la capacité à improviser serait innée, la recherche a démontré qu’elle peut être améliorée par un entraînement délibéré (Brophy, 2001; Kenny et Gellrich, 2002; Kratus, 1991, 1995). Par ailleurs, l’apprentissage et la pratique de l’improvisation exerceraient un effet positif sur l’acquisition et le développement de plusieurs autres compétences, aussi bien musicales que non musicales (Azzara, 1993; Campbell, 2009; Dos Santos et Del Ben, 2004; Kenny et Gellrich, 2002; Koutsoupidou et Hargreaves, 2009; McPherson, 1993; Wilson, 1970). Cependant, l’état actuel des connaissances ne permet pas d’outiller efficacement le pédagogue souhaitant intégrer l’improvisation à la démarche d’apprentissage du musicien classique de niveau collégial ou universitaire (Després, 2011; Després et Dubé, 2015; Dubé et Després, 2012). Par ce projet doctoral, je souhaite contribuer à combler cette lacune en mettant en lumière trois dimensions de l’expertise en improvisation musicale classique : l’acquisition des compétences, la production de l’improvisation et la transmission des savoirs, savoir-faire et savoir-être. Afin de documenter le parcours d’apprentissage, les stratégies de performances ainsi que les approches d’enseignement-apprentissage des instrumentistes et pédagogues experts en improvisation musicale classique, un devis méthodologique en trois phases a été élaboré. La première phase visait à répondre à la question de recherche suivante : « Qu’est-ce qui caractérise le parcours d’apprentissage des improvisateurs experts en musique classique? ». Afin de répondre à cette question, des entrevues ont été réalisées auprès de N = 8 improvisateurs classiques experts de la scène internationale au sujet de leur parcours d’apprentissage de l’improvisation. Ensuite, la deuxième phase visait à répondre à la question de recherche suivante : « Quelles stratégies sont mises en œuvre par les improvisateurs experts en musique classique lors de leurs prestations? ». Une méthode novatrice, reposant sur la stratégie de collecte de données du protocole verbal rétrospectif avec aide à la remémoration subjective, a été mise en place afin de répondre à cette question. N = 5 improvisateurs classiques experts de la scène nationale ont participé à cette phase. La troisième phase de la recherche visait à répondre à la question suivante : « Quels éléments liés à l’expérience, aux représentations, au parcours d’apprentissage ou à la pratique pédagogique des experts du domaine pourraient contribuer à bonifier l’enseignement-apprentissage de l’improvisation musicale classique? ». Au total, N = 15 participants ont été interviewés lors de cette phase. Parmi ces 15 participants, quatre ont été identifiés comme étant experts en improvisation musicale classique, deux comme enseignants experts en improvisation musicale classique et, ces deux dernières catégories n’étant pas exclusives, neuf comme appartenant à la fois à ces deux catégories. Les trois phases de ce projet doctoral ont contribué à enrichir les connaissances au sujet de l’enseignement-apprentissage et de la production de l’improvisation musicale dans le contexte classique, posant ainsi les fondements empiriques d’une pédagogie efficiente de l’improvisation. / Contrary to the widespread belief that the ability to improvise in music is innate, research has shown that it can be enhanced through deliberate training (Brophy, 2001; Kenny & Gellrich 2002; Kratus, 1991, 1995). Furthermore, the learning and practice of improvisation positively influences the acquisition and development of several other musical and non-musical skills (Azzara, 1993; Campbell, 2009; Dos Santos & Del Ben, 2004; Kenny & Gellrich 2002; Koutsoupidou & Hargreaves, 2009; McPherson, 1993; Wilson, 1970). However, the actual state of knowledge is not sufficient to empower the pedagogue wishing to integrate improvisation throughout the learning process of the classical musician at the college or the university level (Després, 2011; Després & Dubé, 2015; Dubé & Després, 2012). This doctoral thesis contributes to the literature by focusing on three dimensions of expertise in classical music improvisation: Skill acquisition, production of improvisation, and transmission of declarative knowledge, skills, and attitudes. In order to document the learning pathways, performance strategies and teaching and learning approaches of expert instrumentalists and pedagogues in Western classical music improvisation, a methodological design in three phases was developed. The research question during the first phase was: What characterizes the learning pathways of Western classical music expert improvisers? To answer this question, interviews were conducted with N = 8 international expert improvisers in Western classical music about their improvisation learning pathways. The second phase aimed to answer the following research question: What strategies are implemented by Western classical music expert improvisers in the course of their performance? An innovative method, based on the retrospective verbal protocol with subjective aided recall data collection strategy was developed to answer this question. N = 5 expert improvisers in Western classical music from the national scene participated in this phase. The third phase of the research was based on the research question: What elements related to the experience, representations, learning pathways, or pedagogical practices of the experts of the domain could help to improve teaching and learning of Western classical music improvisation? In total, N = 15 participants were interviewed during this phase. Among the 15 participants, four were identified as experts in Western classical music improvisation, two as expert pedagogues in Western classical music improvisation and nine as belonging to both categories. The three phases of this doctoral thesis contributed to enriching knowledge about teaching and learning and the production of musical improvisation in the Western classical music context, thus laying the empirical foundations of an effective improvisation teaching practice.