In this paper, we propose a novel method for compressing and optimizing deep neural networks through channel pruning based on spectral norm. As deep learning models grow in complexity, the demand for efficient deployment, particularly in resource-constrained environments, has increased. Traditional pruning methods often rely on magnitude-based criteria, which can overlook the model's stability and generalization capabilities. In contrast, our approach leverages the spectral norm, the largest singular value of a layer's weight matrix, as a more principled criterion for pruning decisions. The spectral norm provides a measure of the network's sensitivity to input perturbations, offering a theoretically grounded method for identifying and removing less critical channels without significantly impacting model performance. We evaluate our method on CIFAR-10 with VGG-16 and ImageNet with ResNet-50, achieving substantial reductions in both model size and computational cost. Our approach leads to a decrease of over 80% in parameters and a 50% reduction in floating-point operations (FLOPs), while maintaining or even improving accuracy compared to traditional pruning techniques. These results demonstrate that spectral norm-based pruning is a robust and efficient method for deep network compression, opening avenues for further research into stability-aware optimization techniques.

Grid-forming (GFM) battery energy storage systems (BESSs) offer a promising solution for improving the dynamic performance and robustness of wind power generation systems. Unfortunately, the inherent coupling between active and reactive power in GFM control degrades the system controllability, making accurate power regulation challenging. This paper systematically investigates the power coupling issue of GFM-BESS in wind farm and introduces a power decoupling control strategy based on feedback linearization. First, a power dynamic response model is established to reveal the power coupling phenomenon. Based on this model, a novel power decoupling control strategy is proposed by compensating the voltage vector reference of the GFM, and its physical essence is further visually explained using phasor diagrams. Then, the power performances of the GFM-BESS for wind farm are investigated by a small-signal model with the proposed decoupling controller. Finally, the validity of the proposed scheme is verified through time-domain simulations.

Grid-forming voltage source converters (VSCs) are regarded as a promising solution for future converter-dominated power systems, but they still demand advanced control schemes to realize their full potential for robust grid operation. This paper presents a decentralized composite control for grid-forming VSC-dominated power systems. The proposed control scheme integrates model predictive control (MPC)-based voltage control with droop-based synchronization, achieving fast voltage regulation and decentralized power sharing. An augmented MPC for voltage control is developed to enable rapid voltage tracking, while mitigating the impact of power/voltage loop coupling and enhancing power regulation. Furthermore, a disturbance observer is incorporated to estimate external unmeasured values and model parameters uncertainties, effectively compensating for disturbances and ensuring accurate control based on local information. The mechanism underlying the performance improvement brought by the proposed controller is investigated through detailed analysis. The validity of the proposed scheme is verified through MATLAB/Simulink and hardware-in-the-loop real-time simulations in both single-and multi-VSC systems.

Basketball is a dynamic sport that requires precise decision-making, often based on instantaneous assessments of player intent. Understanding a player's intention to shoot or pass the ball is crucial for analyzing game strategies, predicting player movements, and enhancing automated commentary. This paper explores the application of logistic regression technique for human intent detection in basketball, specifically focusing on the task of differentiating between shooting and passing movements based on various kinematic features.Human intent detection is a vital aspect of understanding and predicting human actions in various domains, and in sports like basketball, it plays a crucial role in enhancing player performance and game strategies. In basketball, detecting the intent behind a player’s movements and actions can provide valuable insights into their tactics, strategies, and overall game plan. This ability to discern intent from motion data can significantly impact coaching, player training, and real-time game analysis.

A document by Zohaib Akhtar. Click on the document to view its contents.

This paper investigates the effectiveness of digital midterm assessments in enhancing student learning within the context of engineering education. Through a mixed-methods approach, the study combines quantitative analysis of student performance with qualitative insights gathered from lecturer interviews. The aim is to determine whether digital midterms accurately predict final performance, promote deeper student engagement, and contribute to improved learning outcomes. The quantitative results reveal a strong, positive correlation between midterm and final exam performance, particularly for students at the extremes of the performance spectrum. However, the predictive accuracy for students in the middle range is less apparent. Notably, the quantitative analysis primarily provides a descriptive overview and guides the agenda for the semistructured interviews, which form the core of this research. Qualitative findings from the lecturer interviews emphasise the role of digital midterms as both diagnostic tools and motivators for student engagement. The study concludes that digital midterms, when integrated into a balanced assessment strategy, can significantly enhance student learning as evidenced both by student performance and educators' subjective perception of educational value. Key recommendations include refining midterm design, incorporating continuous assessment, and exploring personalised feedback mechanisms. This work contributes to the ongoing discourse on digital assessment practices in higher education, offering practical insights for educators in engineering and related fields.

This paper proposes a method for incorporating decentralized distributed digital (DDD) ledgers, such as blockchain, with multilevel verification. Traditional DDD ledgers like blockchain typically rely on a single verification level, limiting their utility in systems with hierarchical structures where multiple levels of verification are required. In systems where hierarchies naturally arise, integrating hierarchical verification offers a more robust solution. This involves both verifications within individual levels and across different levels, which introduces new challenges. For instance, each level must validate the work of the preceding level, ensuring consistency throughout the hierarchy. This paper addresses these challenges and outlines a framework for tracking the system's state at any given point and assessing its likelihood of failure.

The increasing demand for intelligent services coupled with privacy protection of mobile and Internet of Things (IoT) devices motivates the widespread adoption of Federated Edge Learning (FEL), which enables devices to collaboratively train on-device Machine Learning (ML) models coordinated by the edge server without sharing their private data. However, the intrinsic constraints of device hardware, user behaviors' variability, and the network infrastructure pose significant challenges to FEL, requiring its algorithm design to take into account resource constraints, the need for personalization models, and diverse network conditions. Fortunately, Knowledge Distillation (KD) emerges as an important strategy to address these impediments within FEL. In this paper, we investigate the works that KD applies to FEL, discuss the limitations and open problems of existing KD-based FEL approaches, and provide guidance for their real deployment.

This study explores the probability of God's existence through the Boltzmann brain argument, delving into the roles of fractals and the golden ratio in the universe. By simulating multiple universes, I quantified how often these mathematical phenomena appeared and evaluated their likelihood across different scenarios. The findings showed a notable connection between fractals and the golden ratio, resulting in an estimated probability of God's existence at around 0.9127. This suggests that the complex structures we observe in the universe might point to some guiding intelligence behind their formation. However, I recognized the study's limitations, particularly the reliance on simulated data, which could oversimplify intricate interactions. Future research aims to build on this work by incorporating more advanced models and broadening the simulations to include varied environmental conditions. Overall, this study adds to the ongoing conversation at the intersection of science and philosophy, emphasizing how empirical research can shed light on profound existential questions.

This paper presents a detailed investigation into the performance of multi-level converter topologies (2L-1P-NPC, 3L-1P-NPC, and 4L-1P-NPC) in regulating the angular position and velocity of a switched reluctance motor (SRM). Utilizing the MATLAB-Simulink environment, the study rigorously simulates various operational scenarios under distinct modulation schemes: BVPWM, UVPWM, LSPWM, and VVPWM, all functioning at a constant switching frequency of 10 kHz. Key parameters, including inductances, resistances, and mechanical properties of the load, are meticulously defined to ensure realistic representations of operating conditions. The analysis encompasses both steady-state and transient responses, highlighting the converters' ability to minimize total harmonic distortion (THD) in voltage and electromagnetic torque outputs. Specifically, topology C exhibited the lowest voltage THD of 23.66 % under LSPWM, underscoring the advantages of higher-level converters. Furthermore, the study investigates the effects of duty cycle variations on harmonic profiles and torque characteristics, revealing significant reductions in torque ripple with advanced modulation techniques. Transient performance assessments demonstrate the system's stability and accuracy, with negligible steady-state errors in angular position and responsive behavior to step changes in reference inputs. These findings affirm the efficacy of proportional-integral (PI) control strategies in achieving precise control of the SRM. The results substantiate the superiority of multi-level converter topologies in high-performance applications, providing compelling evidence of their enhanced harmonic performance and control capabilities. This research contributes to the growing body of knowledge in electric motor drive systems, offering insights into the implementation of advanced modulation strategies and the potential for future innovations in converter technologies.

In the domain of continuous control, deep reinforcement learning (DRL) demonstrates promising results. However, the dependence of DRL on deep neural networks (DNNs) results in the demand for extensive data and increased computational complexity. To address this issue, a novel hybrid architecture for actor-critic reinforcement learning (RL) algorithms is introduced. The proposed architecture integrates the broad learning system (BLS) with DNN, aiming to merge the strengths of both distinct architectural paradigms. Specifically, the critic network is implemented using BLS, while the actor network is constructed with a DNN. For the estimations of the critic network parameters, ridge regression is employed, and the parameters of the actor network are optimized through gradient descent. The effectiveness of the proposed algorithm is evaluated by applying it to two classic continuous control tasks, and its performance is compared with the widely recognized deep deterministic policy gradient (DDPG) algorithm. Numerical results show that the proposed algorithm is superior to the DDPG algorithm in terms of computational efficiency, along with an accelerated learning trajectory. Application of the proposed algorithm in other actor-critic RL algorithms is suggested for investigation in future studies.

Pattern recognition-based myoelectric control is traditionally trained with static or ramp contractions, but this fails to capture the dynamic nature of real-world movements. This study investigated the benefits of training classifiers with continuous dynamic data, encompassing transitions between various movement classes. We employed both conventional (LDA) and deep learning (LSTM) classifiers, comparing their performance when trained with ramp data, continuous dynamic data, and continuous dynamic data augmented with a self-supervised learning technique (VICReg). An online Fitts' Law test with 20 participants evaluated the usability and effectiveness of each classifier. Results demonstrate that temporal models, particularly LSTMs trained with continuous dynamic data, significantly outperformed traditional approaches. Furthermore, VICReg pretraining led to additional improvements in online performance and user experience. Qualitative feedback highlighted the importance of smooth, jitter-free control and consistent performance across movement classes. These findings underscore the potential of continuous dynamic data and self-supervised learning for advancing sEMG-PR-based myoelectric control, paving the way for more intuitive and user-friendly prosthetic devices.

This paper explores the intricate effects of wind power integration on the Available Transfer Capability (ATC) of power systems, emphasizing the significance of spatiotemporal correlations in wind speed. We present an innovative optimal power flow model that integrates the Primal-Dual Interior Point Method (PDIPM), ensuring both computational accuracy and efficiency. This research pioneers a systematic analysis of how spatiotemporal wind speed correlations influence wind power output, thereby refining ATC calculations and improving prediction reliability. Furthermore, we assess the impacts of wind farm integration capacity, location, and connection methods on ATC, offering valuable insights for power system planning and market operations.

Language models have achieved remarkable success in generating coherent and contextually appropriate text, but they continue to face significant challenges when applied to tasks that require formal logical reasoning. Introducing structured guidance into the training process allows models to transcend the limitations of purely statistical approaches, providing a systematic means of enforcing logical consistency across multiple reasoning steps. This paper presents a novel approach to improving logical reasoning through program-guided learning, where predefined logical programs guide the model in applying formal logic operations such as conjunction, disjunction, and quantification. The modified Llama model, integrated with a program-guided learning framework, demonstrates significant improvements in task accuracy, consistency, and error reduction across various logical reasoning benchmarks, including symbolic logic, mathematical reasoning, and program synthesis. Experimental results reveal that the enhanced model outperforms the baseline in both accuracy and efficiency, while also showing greater resilience in handling ambiguous tasks through probabilistic logic mechanisms. The approach outlined in this paper contributes to the broader understanding of how language models can be adapted to tackle complex reasoning tasks more effectively, offering new insights into structured learning methodologies.

Mobile Crowdsensing (MCS) is gaining attention for large-scale sensing that involves three types of entities: task requesters, workers equipped with sensing devices, and the platform that assigns tasks to workers considering their objectives and constraints. However, finding an allocation solution that satisfies the conditions above is NP-hard. A few studies suggested approximate solutions to this problem, focusing on one of the task's objectives: coverage maximization. Yet, they implement it in a single-task environment or with weak objective consideration, i.e., they consider other objectives, reducing the utility the task will receive. This study proposes a task allocation that focuses only on maximizing the task coverage, where we improved the solution to consider future task coverage possibilities. We consider an opportunistic MCS environment in which sensing has no impact on user trajectories. We assume a one-to-many matching where a task can be assigned to several workers, while a worker can be matched to at most one task. We first formulate the problem mathematically and prove it to be NP-hard. Then, we design three heuristic-based solutions that are more efficient and perform extensive performance evaluations based on a real-world dataset. Each solution improves the data quality and has a maximum execution time of milliseconds.

Business, academia, and government all these sectors are interrelated, because they keep records of specific types of information. As a result, analyzing data sources from those sector, various databases can be corelate and make prediction. Also, those sectors can benefit by sharing their data with each other. All these data can be analyzed with the help of Artificial intelligence tools and distributed system should be used to run the entire system.

This article introduces a novel strategy for addressing health insurance fraud through the implementation of Graph-Based Social Network Analysis (SNA). By capitalizing on the interconnectedness of coverage holders, healthcare practitioners, and relevant entities, our approach generates an all-encompassing graph structure that encapsulates the complex network dynamics inherent in the health insurance ecosystem. An anomaly detection algorithm and the meticulous construction of node and edge attributes to identify minor patterns are employed to detect fraudulent activity. The outcomes illustrate the effectiveness of the suggested methodology and establish its superiority in comparison to the available alternatives. The contributions of the study to approaches for detecting health insurance fraud enhance the cost-effectiveness and integrity of the system. In light of the persistent gravity of health insurance fraud, this study offers practitioners and regulators in the industry a comprehensive and pertinent response.

Ransomware attacks have become increasingly sophisticated, targeting critical systems and encrypting vast amounts of data, often causing significant financial and operational damage. To address this escalating threat, a novel approach has been developed that leverages feature optimization techniques to enhance the performance of machine learning models for ransomware detection. Through the application of Recursive Feature Elimination (RFE) and Principal Component Analysis (PCA), the study demonstrates that reducing the feature space not only improves detection accuracy but also reduces false positives and computational overhead. The experiments conducted show that optimized models generalize effectively across a variety of ransomware strains, improving both efficiency and scalability. The findings demonstrate the critical role of feature selection in building adaptive and resilient ransomware detection systems, with potential applications in real-time cybersecurity infrastructure.