While ultimate strength analysis methods exist for steel marine structures, there remains a critical gap in understanding the probabilistic behavior of cracked aluminum stiffened plates, particularly in high-speed craft where operational uncertainties compound material and damage variabilities. This gap is addressed through the development of a novel probabilistic framework for analyzing the ultimate strength of cracked aluminum stiffened plates in high-speed craft by integrating nonlinear finite element analysis (NLFEA) with Non-Intrusive Chaotic Radial Basis Function (NICRBF). The framework systematically accounts for uncertainties in crack parameters (size, location, orientation) and material properties, while incorporating operational factors such as wave-induced loads and slamming effects. Through 10000 numerical simulations, it was demonstrated that crack length significantly influences ultimate strength variability (COV ranging from 0.268 to 0.353), while crack orientation affects mean ultimate stress between 253.92-295.73 MPa. For operational conditions, probability of failure increases markedly with vessel speed, reaching 0.994 at design velocity under 8m wave heights. The NICRBF approach provides an efficient computational framework while maintaining accuracy. These findings provide quantitative guidance for structural reliability assessment and safety-oriented design of high-speed aluminum vessels, particularly in determining operational limits based on wave height and speed combinations. The developed framework offers a practical tool for marine engineers to evaluate structural integrity under uncertain damage scenarios.

Black-box attacks play a crucial role in identifying vulnerabilities in deep neural networks by generating adversarial examples that remain imperceptible to human observers. Transfer-based attacks, which leverage surrogate models to craft adversarial examples for attacking victim models, have become a dominant technique in black-box settings due to their practicality and high effectiveness. However, as deep learning models continue to grow in complexity, the resulting disparities between different architectures make generating adversarial examples with strong transferability increasingly difficult. To address this challenge, we propose a novel hybrid query-transfer method called Dual Generalization Attack (DGA). Specifically, DGA enhances model generalization by incorporating a Model Filter (MF) that injects posterior information from the victim model through a limited number of queries, thereby reducing the gap between the surrogate and victim models. Additionally, DGA introduces a Multi-granularity Local Shuffle (MLS) technique to improve input generalization, enhancing input diversity while preserving consistency with the original images. This dual approach reduces the dependency of adversarial examples on both surrogate models and input configurations, thereby enhancing transferability and attack performance on unknown victim models. Extensive experiments demonstrate that DGA significantly outperforms existing black-box attacks, achieving a substantial improvement in adversarial transferability.

Contribution: The purpose of this study is: 1) to identify students' early misconceptions in problem understanding and the use of variables/sequencing in a first-year computer programming course using Python; 2) to quantify how early misconceptions evolve over time, and 3) to characterize how misconceptions in the use of variables/sequencing affect learning more advanced computer programming concepts. Background: It has not been clearly established how early misconceptions evolve over time and how they affect learning later in the semester. Research questions: 1) Which are the most common misconceptions in the use of variables/sequencing in a first-year Python course? 2) Do those early misconceptions persist late in the semester? 3) Does the persistence of misconceptions in variable/sequencing hinder progress in learning more advanced computer programming concepts? Methodology: We assessed student performance (n=116) in multiple choice and short answer questions that involved reading and/or writing small flowcharts or snippets of Python code. Findings: Problem understanding remains a challenge for ~45% of students until the end of the semester. Misconceptions in variable assignment/sequencing persist late in the semester for ~20% of students hindering their progress in learning more advanced computer programming concepts.

Large language models (LLMs) increasingly serve as interactive healthcare resources, yet user acceptance remains underexplored. This study examines how ease of use, perceived usefulness, trust, and risk perception interact to shape intentions to adopt DeepSeek, an emerging LLM-based platform, for healthcare purposes. A cross-sectional survey of 556 participants from India, the United Kingdom, and the United States was conducted to measure perceptions and usage patterns. Structural equation modeling assessed both direct and indirect effects, including potential quadratic relationships. Results revealed that trust plays a pivotal mediating role: ease of use exerts a significant indirect effect on usage intentions through trust, while perceived usefulness contributes to both trust development and direct adoption. By contrast, risk perception negatively affects usage intent, emphasizing the importance of robust data governance and transparency. Notably, significant non-linear paths were observed for ease of use and risk, indicating threshold or plateau effects. The measurement model demonstrated strong reliability and validity, supported by high composite reliabilities, average variance extracted, and discriminant validity measures. These findings extend technology acceptance and health informatics research by illuminating the multifaceted nature of user adoption in sensitive domains. Stakeholders should invest in trust-building strategies, user-centric design, and risk mitigation measures to encourage sustained and safe uptake of LLMs in healthcare. Future work can employ longitudinal designs or examine culture-specific variables to further clarify how user perceptions evolve over time and across different regulatory environments. Such insights are critical for harnessing AI to enhance outcomes.

For many space missions, it is important to estimate the position and orientation of the satellite for operations such as docking and debris removal. It involves the following stages of object detection, landmark regression and pose estimation. For objection detection, we used Faster-RCNN with HRNet as the backbone, the landmark regression part is done using AHRNet architecture, the pose estimation is implemented using PnP algorithm. Firstly, each image was labeled for object detection with bounding boxes around the satellite images created in Blender which were then used to train for satellite detection. An AHRNet was further trained for landmark regression using a 4fold cross-validation approach which involved splitting the dataset into multiple training and validation sets to enhance the Intersection over Union (IoU) metric. After the landmark regression provided a 2D projection of 3D ground truth points, the PnP algorithm was then used for pose estimation. To improve pose estimation accuracy, we integrated solvePnP with an iterator argument utilizing the Levenberg-Marquardt (LM) method to reduce noise and outliers. Our methodology significantly enhances the precision and efficiency of the translation and rotation error of the satellite during the docking processes offering a viable solution for autonomous space missions with potential for future improvements in domain adaptability through the development of unsupervised domain adaptation models. The results of the landmark regression using AHRNet shows an improved reprojection error, orientation and translation errors.

Monolithic integration of III-V compounds onto silicon (Si) substrates utilizing nano-ridge (NR) engineering holds great potential for developing optoelectronic devices that leverage the well-established CMOS process technology. To realize optimized processes that ensure high-quality growth of such novel three-dimensional structures, where feature sizes are in the nanometer range, reliable and rapid wafer-level monitoring is imperative. In this work, we develop an automated and scalable deep learning framework for intricate defect classification in NR structures using scanning electron microscopy images. The proposed method seeks to replace conventional, labor-intensive and error-prone manual defect inspection, significantly reducing turnaround time (TAT) and costs while improving classification accuracy. Furthermore, a graphical user interface (GUI) has been developed to enable efficient deployment in a 300 mm CMOS production environment.

We describe a method for the creation of a compact model for local layout effects (LLE) using pixelated images of the layers of physical layouts as input features. We incorporate these together with electrical measurements of devices on an integrated circuit as input to a neural network. The neural network combines a variational autoencoder (VAE) with a regression network and is used to identify important features of the layout. Then, using a parallel feature neural network to model the effects of those features, we can determine the functional dependence of those effects. Finally, we incorporate those functions into our existing compact model. Because the VAE takes the full design as input, it represents every possible relevant feature. The model visually reveals which features are important, and these visual features can be converted to numerical features. Existing compact models rely on engineers to determine a priori which features are important, but important dimensions on novel devices can easily be missed. We also show that the constraints of the latent space of the VAE allow us to determine during inference whether a particular device geometry is within the domain of the training data. In addition, we describe a process for training the VAE that improves its ability to interpolate more accurately.

The integration of artificial intelligence (AI) into Human Resource Management (HRM) has brought about significant advancements, but it also raises important ethical and social considerations. This paper examines the integration of AI into organizational structures, emphasizing the interplay between technological advancements and human adaptability. Highlights key challenges, including the skill mismatch crisis, leadership shifts, and employee engagement, while presenting a framework for embedding AI strategically through hybrid Human+AI models. The paper outlines a pathway to create inclusive, resilient and AI-enabled organizations that thrive in the evolving workplace landscape. In addition, a use case for AI-based generative software tools for talent management is discussed.

Precision agriculture sensing technologies offer new opportunities for crop monitoring and farming strategy enhancement. However, limited connectivity in rural areas makes sensor data collection challenging. Unmanned Aerial Vehicles (UAVs) have emerged as an efficient alternative for data acquisition in agriculture. Prior research focused on optimizing system performance metrics like energy and flight time, while overlooking the operational costs demanded from farmers, a significant barrier to adopting precision agriculture technologies. This work proposes a novel UAV-assisted agricultural data collection framework for high-precision crop monitoring while reducing UAV operation costs. Specifically, we formulate an Optimal Sensor Set Selection Problem (OSSP) and demonstrate it is NP-Hard. We design an efficient heuristic algorithm called Cost and infeRence Optimization Process (CROP), which exploits data correlation to select the most informative set of sensors within the farmers' budgets. Data collected from these sensors are then utilized to estimate the measurements from unobserved sensors, providing an accurate view of the entire field. We validate our methodology through extensive experiments using a sensor-UAV testbed. Results show that our approach reduces the estimation error up to four times and requires 30% less budget to achieve near-zero inference error, compared to two existing methods.

In response to the evolving landscape of commercial electric vehicles (EVs), we are pioneering the integration of a Power TakeOff (PTO) design and assembly on the e-axle of Commercial Vehicles (CVs).This innovative concept addresses the significant challenges faced by commercial EV manufacturers "The need for auxiliary hydraulic operations without the use of additional motors or external power sources". This study explores the integration of a PTO unit on the rear axle carrier housing of vehicle, which harnesses power directly from the transmission unit. By powering a hydraulic pump, this solution enables seamless operation of auxiliary hydraulic functions crucial for various commercial applications, like tipping and lifting functions. As leaders in this emerging field, our ongoing development efforts are focused on advancing this concept to market readiness.

In today's increasingly complex and volatile stock markets, leveraging advanced machine learning and quantitative techniques is becoming indispensable for enhancing trading strategies and optimizing returns. This study introduces a sophisticated Multi-modal framework that combines Deep Reinforcement Learning (DRL) with Algorithmic Trading Signals and Price Forecasts to improve risk-adjusted returns in equity trading. Utilizing the Proximal Policy Optimization (PPO) algorithm within a custom trading environment built on the FinRL library, our approach integrates advanced algorithmic signals-such as moving average crossovers and oscillator divergence-and incorporates enriched price forecasts from Long Short-Term Memory (LSTM) networks. The proposed framework was rigorously evaluated using a diverse set of 29 out of 30 constituent stocks within the Dow Jones Industrial Average (DJI). The empirical results highlight the effectiveness of the Multi-modal DRL approach, demonstrating significant outperformance over traditional benchmarks, with an annualized return of 16.24%, an annualized standard deviation of 17.49%, a Sharpe Ratio of 0.86, and a Sortino Ratio of 1.27. These findings underscore the potential of Multi-modal DRL frameworks to offer consistent, robust performance and contribute to advancing trading strategies in dynamic market environments.

We propose low-cost intelligent reflecting surfaces (IRSs) inkjet-printed on paper substrates to shape electromagnetic (EM) wave-fronts and control reflected wave directions in the millimeter-wave (mmWave) and sub-Terahertz (sub-THz) bands. Unlike conventional IRSs with circuit elements, an FPGA, and a DC supply, our IRSs use inexpensive paper substrates and conductive nanoparticle silver ink, achieving a phase gradient over a 2π period. Simulations and measurements show that our IRSs steer incoming mmWave and sub-THz signals to anomalous reflection angles. For a 0° incidence angle (θi), our 28-GHz-band IRS reflects an incident wave to a 30° anomalous reflection angle, approximately 12 dB higher than a copper plate. Another 100-GHz prototype shows an 8 dB higher anomalous reflection than a copper plate. IRS beamforming efficiency is related to printing resolution and substrate thickness, crucial for realistic wireless communication environments. These IRSs can direct communication signals to non-line-of-sight (NLOS) locations, extending coverage for current and next-generation wireless networks. The low-cost, simple fabrication process supports potential mass production, facilitating widespread adoption in future wireless communication systems like the Internet of Things (IoT) and 6G.

Remote Sensing (RS) image captioning has traditionally relied on specialized models tailored to domainspecific tasks. The emergence of large vision-language models (VLMs) offers a promising alternative due to their versatility across tasks and domains. However, fine-tuning VLMs for specific applications introduces significant challenges, including computational overhead, overfitting risks, and reduced generalization capabilities. To address these limitations, we propose RAGCap a Retrieval-Augmented Generation framework that leverages pre-trained VLMs for RS captioning without the need for fine-tuning. Our approach employs a similarity-based retrieval model (SigLIP) to identify relevant image-caption pairs from the training set. These retrieved examples, along with the test image, are processed by a multi-image capable VLM (Qwen2VL) using a carefully designed prompt structure. This enables the model to generate captions that not only accurately describe the test image but also preserve the domain-specific style. Extensive evaluations on three RS benchmark datasets demonstrate that RAGCap achieves competitive performance compared to finetuned models while offering enhanced efficiency and generalization. Our framework provides a practical and scalable solution, maintaining the versatility of VLMs while effectively adapting to domain-specific requirements. Code will be available at: https://github.com/BigData-KSU/RAGCap

This paper presents a comprehensive cardiovascular modeling framework that integrates partial differential equations (PDEs), computational fluid dynamics (CFD), and advanced optimization techniques to simulate blood circulation with high accuracy. Unlike traditional models, which rely on simplified assumptions, this approach incorporates pulsatile flow dynamics, non-Newtonian blood viscosity effects, and patient-specific data to enhance predictive capabilities. The governing equations account for arterial compliance, shear stress variations, and complex boundary conditions, providing a more physiologically relevant representation of blood flow. Convergence analysis demonstrates that the numerical solutions approach the exact solution with an error function given by E h = ||u h − u|| L 2 + ||p h − p|| L 2 , confirming a significantly improved convergence rate over classical methods. Stability analysis ensures perturbation decay over time, satisfying d dt ||u ′ || 2 ≤ 0, which guarantees the dissipation of disturbances and validates numerical robustness. A major advancement of this work is the integration of patient-specific parameters, enabling tailored simulations for cardiovascular interventions. The model has clinical applications in optimizing ventricular assist devices (VADs), refining surgical techniques, and improving drug delivery strategies. Future work will focus on real-time validation using clinical data and further refinement of computational efficiency to enhance applicability in medical practice.

This paper investigates network slicing in radio access networks (RANs) that support multiple virtual operators, focusing on maximizing operator utility and ensuring service isolation on a shared physical platform. Network slicing enables efficient resource management and allows for flexible scaling of services based on diverse operator requirements. We evaluate RAN resource slicing for virtual operators by addressing two key challenges: resource reservation and allocation. Operators predict future resource demands using artificial neural networks and reserve a portion of physical resources accordingly. Following reservation, the allocation process is modeled as a Markov decision process aimed at maximizing revenue, which is solved using reinforcement learning algorithms. This dual optimization approach ensures utility maximization and efficient service delivery for virtual operators while balancing network load and enhancing system robustness in dynamic environments. Additionally, our approach accounts for fluctuating user traffic patterns and varying operator priorities, demonstrating significant potential for improving resource utilization, maintaining service-level agreements, and fostering scalable, resilient network operations in highly virtualized environments. Through comprehensive simulation, we establish that the proposed method leads to a significant improvement in resource utilization and operator utility across various traffic loads and operator priority scenarios, clearly demonstrating its competitive advantage in the context of resource reservation and allocation for network slicing in virtualized environments.

This work employs a new approach to analyze coupled-resonator circuits to design and manufacture a fully metallic dual polarization frequency selective surface (FSS). The proposed filtering structure is composed of a series of unit cells with resonators fundamentally coupled along the z-direction and then repeated periodically in the xy-plane. The fully metallic cascaded unit cell is rigorously analyzed within an infinite periodic environment as a coupled-resonator electromagnetic (EM) circuit. The convenient design of the EM resonators makes it possible to push the evanescent EM field through the metallic structure in the desired frequency band for both polarizations. An FSS prototype is manufactured and measured, and good agreement is found between the simulation results and the final prototype.

This brief presents a novel monolithic ultra-low leakage current sensor with large dynamic range. The sensor features an integrated digitizing circuit that generates digital output corresponding to changing leakage current levels, eliminating the need for an explicit ADC. Fabricated in a 180nm BCD process, the sensor occupies just 0.0234 mm 2 of area, consumes 3.1 mW of power and operates at 3.125MHz. It achieves a dynamic range of 25 pA to 5.5 µA, outperforming previous publications by several orders of magnitude. Measurement results demonstrate the sensor's accuracy, with INL and DNL remaining below 0.4 LSB across a temperature range of −40 • C to 125 • C. The sensor's ultra-low sensing capability and wide dynamic range make it suitable for time-dependent dielectric breakdown (TDDB) characterization and in-field reliability monitoring of circuits, enhancing reliability assessments in safety-critical systems.

Non-contact manipulation of floating objects at the air-liquid interface holds significant potential applications in microrobotics, non-invasive microassembly, and lab-on-a-chip systems. However, achieving simultaneous position and orientation (pose) control of floating objects remains a considerable challenge. Here, we present the Vision-based Adaptive Laser Gripper (VALG) system to address these challenges. A modelfree pose control strategy based on equidistant contour scanning lasers (laser grippers) that leverage distributed thermocapillary flow induced by patterned laser scanning is proposed. The adaptive laser gripper can be generated within seconds and relies solely on visual recognition, making it applicable to a wide range of floating objects without the need for prior modeling or additional information. The effectiveness and control precision of the VALG system is experimentally validated through discrete pose verification and continuous pose control tests. The maximum position and orientation errors are 0.05 mm and 2.6 • for discrete pose verification, and 0.4 mm and 14.5 • for continuous pose control, respectively. The VALG system enables smooth, precise, fast, and adaptive pose control of floating objects, establishing it as a universal and versatile platform for noncontact manipulation at the air-liquid interface.