22505 research outputs found
Sort by
‘Reflective Learning’ – Teaching the South African Military Academy Way
No full textProfessional military education for young officers within the South African National Defence Force (SANDF) has since 1950 been vested within the South African Military Academy (SAMA), Faculty of Military Science, Stellenbosch University. As an all services academy, the SAMA houses student from all four arms of service that constitute the SANDF, namely the South African Army, Navy, Air Force and Medical Health Services. With such a vastly different student body, the five schools within the SAMA correspondingly cover an extremely varied field of study ranging from military law and military management to military psychology all the way to more traditional military subjects such as military history and strategy.
The question must thus be asked – what teaching approach or teaching philosophy should the Military University Educators (MUE) of the SAMA utilise in educating their learners?
The answer, this paper will argue, lies in a reflective learning philosophy that is tailored to meet the requirements of the South African armed forces. This reflective teaching philosophy, shaped for the military environment, involves a thoughtful and iterative approach to instruction that emphasises critical thinking, self-awareness, and continuous improvement. This philosophy also recognises that military professionals operate in dynamic and complex environments where adaptability and the ability to learn from experience are crucial. A skill that the SAMA intends fostering within its student body of young military officers during their time spent at the SAMA
The Potential Impact of Simulation Based Learning in the Kosovo Security Force
No full textAs the Kosovo Security Force (KSF) undergoes a strategic transformation into a fully operational military force, the need for modern and adaptive training methods has become increasingly urgent. While recent investments have focused on personnel growth, weapon procurement, and defense infrastructure, the integration of simulation-based learning (SBL) into training programs has not yet been considered. This study explores the potential of SBL - particularly virtual reality (VR), augmented reality (AR), and wargaming - as a tool for enhancing the readiness and adaptability of KSF personnel. Drawing on existing literature, the study evaluates the benefits and challenges of each technology in relation to the specific context of Kosovo. Findings suggest that while VR and AR offer immersive and technologically advanced experiences, their high cost and implementation complexity may limit immediate applicability. Wargaming, by contrast, presents a practical and cost-effective entry point for introducing SBL, with proven benefits in improving decision-making, stress management, and strategic thinking. The study concludes that integrating wargaming into KSF’s training programs would support their modernization goals and better prepare personnel for the complexities of modern warfare
Exploring Human Perception and Cognition in Expressing and Understanding Mechanical Designs
No full textEffective communication of mechanical designs through technical drawings requires geometric accuracy, efficiency, and an understanding of human perception and cognition. Although advances in computer-aided design (CAD) software have improved drawing precision and automation, current tools often overlook the spatial reasoning processes that users employ to interpret these representations. This study seeks to improve the accuracy and efficiency of human-computer interaction in CAD environments by examining how individuals perceive and interpret three-dimensional mechanical components within the context of technical drawings. A key focus is the identification of canonical and optimal views that align with intuitive human understanding. Through a series of four experimental studies, this dissertation breaks down the optimal view selection into four core dimensions (steps): (1) orientation, (2) rotation, (3) visibility of features, and (4) display/viewing format. The findings demonstrate that users consistently prefer views with upright vertical alignment, a high number of visible effective edges, and minimal reliance on unimaginable or occluded features. These preferences are grounded in psychological theories of spatial cognition, human factors research, and established engineering design standards. In contrast to AI-driven methods that rely solely on pattern recognition to infer optimal views, this study introduces a human-centered framework that can be integrated into intelligent CAD systems. By bridging engineering design with perceptual science, this research supports the development of smart CAD tools that enhance the accuracy, efficiency, and usability of technical communication in mechanical design
Out To Pasture
No full textWith a wildly successful acting career approaching its final days, Marcel T. Cow looks back at his rise to fame, as well as a rollercoaster of a career. A toxic relationship, substance abuse, and living a lavish lifestyle all lead to his downfall, ultimately making him face one final decision: Leave the vices behind and save his career, or give in and lose everything. Out To Pasture is an animated biopic parody about Marcel’s life as an actor and how his career comes to an end. The film starts with a flashback to Marcel’s humble beginnings, moving to the Big City for the first time and getting his first gig on a sitcom after giving the audition of a lifetime to a producer who just so happened to be looking for a lead. Starting his career off strong, Marcel continues his rapid journey to success. He starts to live the high life, meets a boy, and parties hard. But before he can stop himself, he finds himself facing substance abuse, with no intention of finding help. It negatively impacts his career and leads to a messy breakup, making him turn to the only thing he knows, drugs. In the end he’s presented with a choice, to continue his downward spiral or to save his dying career
Three Forks
No full textThree Forks is a thesis drama series about a 19th-century woman sent West to live with her estranged family in Muscogee, a lawless town where justice bends to power and survival means learning the game — or being buried by it. As we follow our lead, Eleanor Davis, out West and eventually her settlement into town, we gain a look into her new world, one where tribal council contend with encroaching U.S. jurisdiction and where settlers, freedmen, and railroad men jostle for influence in a town on the brink of statehood. The many fabled stories and myths about the West she grew up with colliding with the messy truth her journey reveals. With this, Eleanor represents the struggle between the old and new, between East and West and while the show highlight the violence that happened in this period of time within the Indian territory and those displaced because of it, the overarching storyline within Three Forks is about a place that is falling into a new century with a promise of brighter horizons, but leaving people to wonder: for whom? Those that know how the play the game or those that recognize that the rules were never guaranteed to stay the same
Advancements in ML via Efficient Generative Modeling, Robust Domain Adaptation, and Explainable Multimodal Retrieval
No full textThe rapid evolution of AI heightens the need for learning systems that are efficient, robust, and explainable. This dissertation advances these three pillars through innovations in classification, generative modeling, domain adaptation under data-scarce conditions, and multimodal retrieval. Collectively, the methods reduce dependence on large, labeled datasets, improve adaptability under distribution shifts, enable deployment on resource-constrained platforms, and enhance interpretability. For classification, the Iterative Maximum Likelihood Classifier (IMLC) recasts regularized maximum likelihood training as a fixed-point contraction with convergence guarantees, enabling faster and more stable optimization. Results on synthetic data and MNIST validate its efficiency. For generative models, we introduce Parametric Mish (PMish) activation, a modified MMD-GAN repulsive loss, and Adaptive Rank Decomposition (ARD) for compression. Together, these stabilize training, improve convergence, and reduce model size while maintaining fidelity. Benchmarks (CIFAR-10/100, STL-10, CelebA) show strong image quality with lower resources. Moving from model efficiency to data robustness, two data augmentation methods, namely Shuffle PatchMix (SPM) and Dual-Region Augmentation (DRA), are proposed to improve model robustness in data-scarce scenarios. Using these augmentations, along with novel loss functions and confidence-margin reweighting for noisy pseudo-labels, yields consistent gains under distribution shift. For multimodal retrieval, X-CoT combines coarse video-text matching with large language model chain-of-thought reasoning to raise retrieval accuracy and produce human-interpretable rationales. Together, these contributions demonstrate that performance, adaptability, and transparency can be achieved, laying a foundation for next-generation machine learning systems that are efficient, robust, and explainable
Practice?
No full text3D animation is a combination of art and technique. Art serves to convey your perspective, express beauty, tell stories, and more. However, to accomplish these goals, we need solid technical support. Only by paying attention to both, can we achieve excellent results. The title of this film is “Practice?”, which is a deliberate pun. On one hand, practice refers to the martial arts training sequences of the character, during which his master provides guidance, intervenes in his movements, and progressively increases the difficulty. On the other hand, practice also symbolizes my own artistic and technical experimentation—exploring and refining skills in rigging, VFX, and other aspects throughout the process. To avoid making the content overly technical or monotonous, I incorporated a master character who guides or interferes with the protagonist’s actions through voiceover, weaving these elements into an absurdist comedy. This narrative device connects the technical demonstrations into a cohesive storyline
Rise of Social Media Hacking: AI-Based IP Tracking for UAE Law Enforcement
No full textSocial media has evolved into a critical channel for communication, expression, and public influence, but it has also become a prevalent avenue for cybercrime, particularly in digitally advanced nations such as the United Arab Emirates (UAE). The rising complexity of online offences, coupled with anonymisation tools and cross-border digital behaviour, has made the attribution of social-media-based cyber incidents increasingly challenging for law enforcement. In this context, artificial intelligence (AI) offers the potential to strengthen digital investigations by providing intelligent, scalable, and evidence-driven attribution capabilities. This research develops an AI-assisted Internet Protocol (IP) attribution framework tailored specifically for UAE law enforcement needs. The study addresses a significant gap in existing global models, which often overlook the UAE’s legal, cultural, linguistic, and infrastructural considerations. A fully synthetic dataset was generated for this research to ensure zero privacy or compliance risk. Machine learning techniques and forensic analytics were examined to determine their effectiveness in analysing behavioural patterns, device indicators, and network signals associated with suspicious online activity. The study further proposes a UAE-aligned system blueprint, designed to support digital evidence strengthening, enhance attribution confidence, and enable responsible operational use by cyber investigation units. The findings demonstrate that AI can meaningfully contribute to proactive cybercrime detection, attribution, and forensic intelligence when paired with regulatory alignment and contextual system design. Beyond technical validation, the thesis provides a strategic roadmap for integrating AI-based IP attribution into existing UAE law enforcement workflows, outlining adoption phases, governance requirements, and policy considerations. The research offers a foundational, ethical, and implementation-ready framework that can be extended through future real-world data collaboration and system-level deployment. Keywords: Social media cybercrime, Artificial Intelligence, IP attribution, Digital forensics, UAElawenforcement, Cybersecurity, Forensic analytic
Advancing Applications of Deep Learning on Network Traffic Analysis and Fingerprinting
No full textThe growing reliance on encrypted communication networks and privacy-preserving technologies, such as Tor, has intensified the demand for advanced defenses against traffic analysis attacks. Although encryption conceals content, the exposure of metadata—such as timing, packet size, and traffic volume—remains a significant vulnerability, allowing adversaries to infer private user behavior and visited websites. This dissertation addresses the limitations of existing analysis by advancing the application of deep learning techniques across multiple domains of traffic analysis and fingerprinting. The first major contribution is a comprehensive evaluation of current website fingerprinting defenses, revealing critical weaknesses against sophisticated modern attacks. Building on this foundation, a novel attack is introduced, utilizing transformer architectures and enhanced feature representations to capture complex, long-range dependencies in network traffic and significantly improve attack accuracy against defenses. Additionally, this dissertation presents an improved flow correlation attack on Tor, leveraging a transformer-based model to more effectively correlate traffic between entry and exit nodes, thus heightening the threat to user anonymity. Finally, this dissertation demonstrates an adaptation of these traffic analysis techniques for defense in the form of stepping-stone intrusion detection, addressing key challenges such as protocol variability, multi-hop complexity, and traffic obfuscation to develop more robust and adaptable detection methods in complex network environments
