1393 research outputs found
Sort by
Adaptive and Scalable Edge-Based Weed Classification Through Distilled Mixture of Experts And Metric Learning
No full textModern agricultural vision systems must identify weeds reliably while working with very little labeled data and under strict limits on compute and memory. This dissertation introduces a few-shot weed recognition framework that combines a mixture of experts student with a strong teacher through knowledge distillation and task-conditioned routing. The teacher is a partially fine-tuned EfficientNet-B7 that produces high-quality embeddings. The student consists of lightweight expert backbones drawn from MobileNetV3, ShuffleNetV2, and EfficientNet-B0. A routing gate summarizes the support set of each episode into a compact context vector and uses that context together with image features to select the most suitable experts through top-K sparse routing.
Training proceeds in two stages. A short distillation warm-up first aligns the student embeddings with the teacher. The system then adds metric learning with a semi-hard triplet objective driven by a cosine-margin schedule. The implementation uses automatic mixed precision and true gradient accumulation and is designed to be numerically safe for floating point FP16. For deployment, the trained student supports post-training dynamic quantization to INT8 for efficient CPU inference.
Evaluation follows standard N‑way, K‑shot episodic protocols. In the primary 5‑way, 5‑shot setting on our eight‑class weed benchmark, the model achieves 97.13% ± 0.57 top‑1 episodic accuracy averaged across three independent training seeds (1,000 episodes per seed). Accuracy is preserved after INT8 quantization, while stored model size is reduced from 46.86 MB to 29.90 MB. Comprehensive ablations isolate the contribution of each component, including removal of triplet mining, removal of distillation, task-conditioned routing turned off, and a frozen teacher without partial fine-tuning. Additional studies compare uniform routing without a gate, and a single expert baseline.
The results show that task-conditioned routing is accuracy-neutral relative to image-only and uniform routing controls on the core dataset, indicating that most gains come from the quality of the distilled embedding and prototype-based inference rather than the specific gating policy. Combining distillation with metric learning consistently outperforms either component in isolation; removing distillation or triplet mining leads to notable drops in few-shot accuracy, confirming that the two objectives are complementary. Distillation transfers class-structure from the teacher while triplets sharpen inter-class margins. Quantization to INT8 preserves accuracy within the confidence bounds of episodic evaluation while reducing model size by approximately 36% and improving CPU throughput, which is critical for edge deployment. Beyond the base benchmark, the same trained artifact adapts to novel species without weight updates: in episodic evaluation on previously unseen PlantSeedlings classes, the model attains 97.01% mean accuracy (95% CI ± 0.12 percentage points (pp)) under a 5-way, 5-shot, 15-query protocol and in bank-based deployment across DeepWeeds and PlantSeedlings it achieves a cosine mean accuracy of 93.34%. Taken together, these findings provide a practical path to accurate, robust, and efficient few-shot weed recognition that supports forward-only adaptation in resource-constrained field conditions
Utilizing Ensemble Learning Techniques to Enhance Corn Price Prediction: A Case Study on South Dakota
No full textPredicting crop prices is a complex challenge that farmers must navigate each year, but machine learning algorithms can provide valuable insights to support more informed decision making. In recent years, agricultural price prediction models have made significant advances, with architectures achieving varying degrees of success. However, ensuring the accuracy and reliability of these models remains an ongoing challenge. This research explores the use of stacking, an ensemble learning technique, to enhance the performance of base models in predicting corn prices in many regions of the state of South Dakota in the USA. We propose a hybrid architecture that combines Long Short-Term Memory Networks and Transformer Neural Networks, allowing us to leverage the strengths of both models. Using real corn price data from the last 11 years, our findings show that this stacked architecture not only outperforms its individual base models but also industry standard approaches. The ensemble model can predict within a one-month window, reaching a maximum mean average error of 6%. More research is needed to use multiple features, such as weather, crude oil price, and market demand, to provide an efficient decision support system
Decentralized Multi-Hop Federated Reinforcement Learning for Energy-Efficient and Secure Routing in LoRaWAN-based Smart City Infrastructure
No full textThis paper proposes a decentralized hybrid framework that integrates Federated Learning (FL) and Reinforcement Learning (RL) to enable energy-efficient and secure multi-hop routing in LoRaWAN-based smart city networks. The method allows local model training at gateways and relay nodes, combining real-time routing decisions with privacy-preserving federated aggregation. Key metrics such as residual energy, link quality, and node trust levels are embedded into the reward function, and lightweight encryption plus differential privacy safeguard routing metadata. Simulation results demonstrate substantial gains in packet-delivery ratio, latency, network lifetime and resilience to adversarial attacks when compared to traditional routing protocols
Quantifying Return on Controls in LLM Cybersecurity
Get PDFAs large language models are increasingly deployed in real world systems, concerns about their security have grown. While these models demonstrate impressive fluency and utility, they are also susceptible to a variety of attacks, including prompt injection, data leakage, and adversarial manipulation. These vulnerabilities pose serious risks in high stakes environments where privacy, reliability, and trust are essential. Despite emerging efforts to document and mitigate such issues, a gap remains in the ability to systematically measure the effectiveness of proposed defenses in a quantifiable and reproducible manner.
This research addresses the question: how can security vulnerabilities in language model based systems be modeled, measured, and mitigated using a quantitative risk framework? To explore this, a retrieval augmented generation application was constructed using the DeepSeek-R1 language model, FastAPI, LangChain, and FAISS. The application was tested under baseline conditions and then with three security interventions: attribute based access control, named entity recognition redaction using Microsoft Presidio, and NeMo Guardrails for response filtering.
Five vulnerability types were tested: exposure of personally identifiable information, latent context injection, adversarial prompt generation, direct prompt injection, and divergence. A probe based tool was used to simulate adversarial attacks and measure susceptibility. Attack success probabilities were estimated using Laplace’s Rule of Succession, and expected losses were modeled using triangle distributions informed by breach cost data. Monte Carlo simulations were used to project cumulative losses under each configuration, and loss exceedance curves were generated to visualize long tail risk.
The results showed that two of three controls led to measurable improvements in system security. The third control did not provide significant protection out of the box. The access control mechanism effectively reduced attack likelihood but did so by disabling document retrieval entirely. The named entity recognition filter allowed continued use of retrieval augmented generation while reducing exposure to sensitive data. NeMo Guardrails provided no robust post generation filtering out of the box, producing an expected loss comparable to the baseline and the least return on control. This demonstrates the value of layered defenses that minimize both the probability and impact of successful attacks.
This thesis contributes a reproducible methodology for evaluating the security of large language models using probabilistic models, financial risk simulations, and comparative cost benefit analysis. It provides practitioners with tools to assess the effectiveness of different controls and highlights the importance of structured, data driven approaches to securing artificial intelligence systems
Quantifying the Effectiveness of Dynamic Response in Web Application Honeypots
Get PDFThis research quantifies the effect dynamic response has on the effectiveness of web application honeypot systems. While much academic effort has been directed toward honeypot development, few scholars have measured the effect of various honeypot tools and compared them against each other. This research measured the effectiveness of honeypot dynamism through a standardized experiment between a Tanner/Snare honeypot system, a Glastopf honeypot system, and a static honeypot system. Using the T-Pot framework, this study analyzed the duration of attacker engagement and their actions across various honeypots. Results show that the Tanner/Snare honeypot engages attackers in a fundamentally different way than the Glastopf honeypot, but this does not translate into a longer attacker engagement time. By contrast, the Glastopf honeypot engages attackers for longer in reconnaissance-based attacks but does not encourage further exploitation. This research assists cybersecurity professionals and honeypot developers in making informed decisions about where best to direct their efforts, improving the overall cyber deception landscape
Impact of artificial intelligence and knowledge management on proactive green innovation: the moderating role of trust and sustainability
No full textPurpose
In this research, we seek to understand the effects of artificial intelligence (AI) and knowledge management (KM) processes in enhancing proactive green innovation (PGI) within oil and gas organizations. It also aims to investigate the moderator role of trust and sustainability in these relationships. Design/methodology/approach
This paper employs a quantitative analysis. Surveys have been gathered from the middle-line managers of twenty-four oil and gas government organizations to evaluate the perceptions of the managers towards AI, KM processes, trust, sustainability measures and proactive measures toward green innovation. Analytical and statistical tools that were employed in this study, including structural equation modeling with SmartPLSv3.9, have been used to analyze the data and to examine the measurement and structural models of this study. Findings
The study results reveal a significant and positive impact of AI utilization, KM processes and PGI within oil and gas organizations. Furthermore, trust and sustainability turn out to be viable moderators affecting, and influencing the strength and direction of AI, KM and PGI relationships. In particular, higher levels of trust and more substantial sustainability commitments enhance the positive impact of AI and KM on green innovation outcomes. Practical implications
Understanding the impact of AI, KM, trust and sustainability offers valuable insights for organizational leaders and policymakers seeking to promote proactive green innovation within the oil and gas industry. Thus, organizations can increase the efficiency of sustainable product development, process improvement and environmental management by using robust AI technologies and effective KM systems. Furthermore, fostering trust among stakeholders and embedding sustainability principles into organizational culture can amplify the effectiveness of AI and KM initiatives in driving green innovation outcomes. Originality/value
This study extends the current knowledge by assessing the effect of AI and KM on proactive green innovation while accounting for trust and sustainability as moderators. Utilizing quantitative methods offers a nuanced understanding of the complex interactions between these variables, thereby advancing theoretical knowledge in the fields of innovation management, sustainability and organizational behavior. Additionally, the identification of specific mechanisms and contextual factors enriches practical insights for organizational practitioners striving for a practical understanding of the dynamics of the complexities of sustainable innovation in an AI-driven era
Adopting knowledge-sharing-driven blockchain technology in healthcare: a developing country’s perspective
No full textPurpose
This study aims to investigate the key factors that influence the behavioural intention of doctors to adopt the knowledge sharing driven blockchain technology in government hospitals. The study is based on the Unified Theory of Acceptance and Use of Technology 2, with the addition of trust as an independent variable and knowledge sharing as a mediating variable between trust and behavioural intention. Design/methodology/approach
The data for the study was collected through a correlation and cross-sectional study using a survey, with a sample of 322 responses being used for the final analysis. The initial analysis of the data was conducted using SPSS v.26, followed by a partial least squares structural equation modelling (PLS-SEM) analysis using SmartPLS v.3.9 to test the validity and reliability of the measures and to examine the hypothesized relationships. Findings
The results supported the proposed framework. The results of PLS-SEM indicate that all proposed pathways support the model. In particular, the results of the study reveal that performance expectation, effort expectation, social influence, facilitation conditions and trust are drivers of blockchain adoption and have a significant impact on the behavioural intention of clinicians in hospitals. Furthermore, the study found that knowledge sharing mediated the relationship between trust and behavioural intention. Practical implications
The present study sheds light on the challenges facing blockchain technology, such as privacy and trust concerns and proposes a more sustainable approach based on knowledge management to enhance the effectiveness of blockchain technology and overcome these challenges. Originality/value
The significance of this paper lies in the limited literature examining the relationships between blockchain technology and knowledge management processes. Furthermore, a hypothetical framework that includes the knowledge sharing process as a mediating variable between trust and behavioural intention to adopt blockchain technology has not been presented or developed in any previous studies, particularly in the context of Iraq. Thus, this work is novel and unique in its approach
Knowledge Types, Mechanisms, and Media in Survey Aviation: Case Study and Implications
No full textEffective knowledge management (KM) is predicated on understanding the knowledge types, knowledge mechanisms, and media (KTMM) that support organizational activities. Our study investigates KTMM in the context of low-altitude aerial surveying, a knowledge-intensive business process that relies on collaboration between aviators and observers to devise and execute procedures safely and efficiently. Through the case study method, we find that aerial surveyors rely primarily upon tacit knowledge and social mechanisms for knowledge capture and sharing. Additionally, we find a preference for rich visual communication media. Our study contributes knowledge in a hitherto unstudied context and provides insights on how KM occurs in a context where heterogeneous agents must collaborate. In addition, our study offers practical implications for KM system design and KM strategies
Pure and Strong Nash Equilibrium Computation in Compactly Representable Aggregate Games
Get PDFAggregate games model interdependent decision making when an agent’s utility depends on their own choice and the aggregation of everyone\u27s choices. We define a compactly representable subclass of aggregate games we call additive aggregate games, which encompasses popular games like congestion games, anonymous games, Schelling games, etc. We study computational questions on pure Nash equilibrium (PNE) and pure strong Nash equilibrium (SNE). We show that PNE existence is NP-complete for very simple cases of additive aggregate games. We devise an efficient algorithmic scheme for deciding the existence of a PNE and computing one (if it exists) for bounded aggregate space. We also give an approximation algorithm for a special type of additive aggregate games. For SNE, we show that SNE recognition is co-NP-complete and SNE existence is \Sigma_2^P-complete, even for simple types of additive aggregate games. For broad classes, we provide several novel and efficient aggregate-space algorithms for recognizing an SNE and deciding the existence of an SNE. Finally, we connect our results to several popular classes of games and show how our computational schemes can shed new light on these games