Historically, grid-connected synchronous machines have formed the grid voltage and frequency. Today, the high penetration of inverter-based resources poses challenges to grid stability, as conventional grid-following control methods do not provide grid-forming capabilities. New grid-forming control methods need to stabilize the grid. Therefore, electromagnetic transient modeling is essential for control design and stability analysis in future power systems. This paper proposes a novel reduced-order modeling and identification approach for doubly-fed induction machines (DFIMs) with a grid-connected stator and inverter-connected rotor. The proposed generic modeling remains valid under varying grid or stator conditions. Consequently, the modeling approach is also applicable to induction machines (IMs) with an inverter-connected stator. In this paper, DFIM measurements identify a holistic currentto-flux mapping to model nonlinear magnetic saturation effects. A virtual current injection method is introduced to identify all differential inductances without additional measurements. Various simplified and holistic nonlinear modeling approaches are compared, and measurements validate the proposed holistic flux dynamics (HFD) model under varying grid conditions.

Chest X-rays are one of the most widely used diagnostic tools in medical imaging, offering critical insights into pulmonary, cardiovascular, and thoracic conditions. The advent of deep learning has significantly enhanced the ability to analyze chest X-rays, enabling accurate and automated detection, classification, and interpretation of a wide range of pathologies. This survey provides a comprehensive overview of state-of-the-art deep learning methods applied to chest X-ray analysis. It categorizes existing approaches based on their primary tasks, including disease classification, localization, segmentation, and radiology report generation, highlighting their strengths, limitations, and clinical applicability. Key challenges, such as data scarcity, variability across imaging systems, lack of interpretability, and integration into clinical workflows, are discussed in detail. The survey also explores promising directions for future research, including self-supervised learning, federated learning, multimodal data integration, and the development of explainable and robust AI systems. By synthesizing recent advancements and identifying critical gaps, this work aims to guide researchers and practitioners toward the development of clinically impactful solutions. The findings underscore the transformative potential of deep learning in chest X-ray analysis while emphasizing the need for multidisciplinary collaboration to address existing barriers and advance the field.

Recent advancements in multi-agent reinforcement learning (MARL) have demonstrated its potential and its applications in modern games. Beginning with foundational work and progressing to landmark achievements such as AlphaStar in StarCraft II and OpenAI Five in Dota 2, MARL has proven capable of achieving superhuman performance across diverse game environments through techniques like self-play, supervised learning, and deep reinforcement learning. With its growing impact, a comprehensive review has become increasingly important in this field. This paper aims to provide a thorough examination of MARL's application from turn-based two-agent games to real-time multi-agent video games including popular genres such as Sports games, First-Person Shooter (FPS) games, Real-Time Strategy (RTS) games and Multiplayer Online Battle Arena (MOBA) games. We analyze critical challenges posed by MARL in video games, including nonstationary, partial observability, sparse rewards, team coordination, and scalability, and highlight successful implementations in games like Roller Champions, Rocket League, Minecraft, Quake III Arena, StarCraft II, Dota 2, Honor of Kings, etc. This paper offers insights into MARL in video game AI and also proposes future research directions to advance MARL and its applications in game technology, inspiring further innovation in this rapidly evolving field.

The possibility to predict the epitope recognized by a given T-cell receptor (TCR), based on the Complementarity Determining Region 3 (CDR3) sequence, could significantly facilitate cancer immunotherapy approaches. However, accurate prediction of TCR peptide binding is still a significant challenge. We have developed a new machine-learning-based framework to support adoptive Tcell therapy for cancer. Our framework predicts the probability of the presence of specific amino acids in the epitopes presented by MHC (Major Histocompatibility Complex) that are recognized by a given CDR3 site in a TCR expressed by a T lymphocyte. We also propose a new intuitive approach to the numerical evaluation of the results. Both the conventional and proposed numerical evaluation of our results show that the algorithm captured the data distribution and predicted epitope chains with reliable accuracy. These findings demonstrate that our machine-learning framework lays the foundation for new opportunities in personalized medicine. Moreover, the approach that we propose has a high level of generality and could be used for any kind of sequence-to-sequence problem where specific positions in the predicted sequences are important.

This Technical paper presents an innovative approach to enhancing cybersecurity by leveraging advanced Network Address Allocation (NAA) techniques. Using dynamic address assignment and automated management systems, the paper demonstrates how organizations can proactively reduce vulnerabilities, enhance network visibility, and improve overall security. The inclusion of automated solutions for identifying available IP addresses in real-time further optimizes the NAA process. This approach has been shown to significantly reduce attack surfaces, improve monitoring, and ensure more resilient industrial networks. By adopting these innovations, industries can stay ahead of evolving cyber threats, ensuring secure and uninterrupted operations in the digital age.

Detection and prediction of the onset of seizures are among the most challenging problems in epilepsy diagnostics and treatment. Small electronic devices capable of doing that will improve the quality of life for epilepsy patients while also open new opportunities for pharmacological intervention. This paper presents a novel approach using machine learning techniques to detect seizures onset using intracranial electroencephalography (EEG) signals. The proposed approach was tested on intracranial EEG data recorded in rats with pilocarpine model of temporal lobe epilepsy. A principal component analysis was applied for feature selection before using a support vector machine for the detection of seizures. Hjorth's parameters and Daubechies discrete wavelet transform coefficients were found to be the most informative features of EEG data. We found that the support vector machine approach had a classification sensitivity of 90% and a specificity of 74% for detecting ictal episodes. Changing the epoch parameter from one to twenty-one seconds results in changing the redistribution of principal components" values to 10% but does not affect the classification result. Support vector machines are accessible and convenient methods for classification that have achieved promising classification quality, and are rather lightweight compared to other machine learning methods. So we suggest their future use in mobile devices for early epileptic seizure and preictal episodes detection.

The transition from idealized descriptions of memristors to physical implementations, such as resistive random access memory (RRAM) devices, is crucial for realizing many neuromorphic computing concepts in hardware. However, the intrinsic variability and fabrication tolerances of real-world RRAM devices present significant challenges for designing reliable circuits. This study focuses on addressing these challenges in the context of a memristive Hindmarsh-Rose circuit, a model for simulating the spiking and bursting behavior of biological neurons. To evaluate the robustness of the circuit in maintaining oscillatory behavior the Variability Intensity Model (VIM) is introduced, which enables a two-stage analysis with respect to the absolute tolerances and the dynamic variabilities. The results show that high conductance states are the most sensitive parameters, while low conductance states are more robust, optimally matching the behavior of real RRAM devices where relative variability decreases with increasing conductance. In addition, the tolerance analysis shows that spiking dynamics are generally robust to parameter variation, while bursting behavior is more sensitive. These results establish parameter constraints for actual RRAM devices as a basis for designing robust RRAMbased implementations of the Hindmarsh-Rose circuit.

This study proposes an innovative methodology for accurate vehicle maneuver analysis using time-differenced GNSS carrier-phase (TDCP) measurements from smartphones, combined with a vehicle motion estimation approach based on factor-graph optimization (FGO). By leveraging the centimeterlevel precision of TDCP, the method achieves accurate vehicle displacement measurements, while the FGO framework enhances the estimation of vehicle speed, acceleration, pitch and heading angles, as well as angular velocity. Extensive experiments involving multiple vehicles and drivers validate the method's effectiveness in vehicle motion estimation, maneuver identification, and risk classification. The results demonstrate a reduction in speed and angular velocity estimation errors by up to 47.28% and 66.95%, respectively, compared to traditional GNSS solutions. Furthermore, classification accuracy for highrisk driving behaviors improves by 27.37%, making the method suitable for applications such as usage-based insurance (UBI), fleet management, and autonomous driving.

This study evaluates the effectiveness of a 6-hour Calculus for Engineering preparation course designed for high school seniors (Group A) and incoming first-year engineering students (Group B), utilizing advanced educational technologies. Unlike the traditional 9-15 hour approach at the Engineering Faculty of Rajamangala University of Technology Krungthep (RMUTK), which begins with foundational topics such as numbers, function values, limits, and continuity, our course focuses directly on differentiation using curve slopes and symbolic mathematics. The course employs GeoGebra and SymPy Live web applications to facilitate learning. Paired sample t-tests revealed significant improvements in both groups: Group A (n = 35) showed a mean increase of 4.97 (t(34) =-10.62, p < 0.001), and Group B (n = 27) showed a mean increase of 2.93 (t(26) =-4.05, p < 0.001). An independent samples t-test indicated a significant difference in improvement between the groups (t(60) = 2.47, p = 0.016). The results suggest that the 6-hour calculus preparation course effectively enhanced the calculus competency of both high school seniors and incoming first-year engineering students, with Group A showing greater improvement than Group B. This provides a more efficient alternative to the traditional methods used at RMUTK.

The advent of Generative Artificial Intelligence (GenAI) has sparked significant interest in education, offering ways to support learning, personalize student experiences, and boost engagement. Generative AI holds the promise of transforming education with personalized learning, instant feedback, and assistance with complex problem-solving. However, its integration into classrooms requires careful management due to ethical concerns, misinformation risks, and potential misuse. While many articles explore the potential of generative AI in education, empirical studies on its real-world classroom use are limited. This paper presents an experience report on deploying a customized GPT-powered chatbot at Qatar University to support learning in two undergraduate courses: Data and Computer Communications Networks (technical) and Computer Ethics. Working across these diverse courses allows a thorough analysis of generative AI's strengths and weaknesses in different academic contexts, offering a comprehensive evaluation of its applicability and effectiveness. We used a mixed-methods approach with over 100 students, collecting quantitative and qualitative data via questionnaires to assess the chatbot's impact. Findings are analyzed with established theoretical frameworks to contextualize the pedagogical impact of generative AI, aligned with UNESCO guidelines for ethical integration. This paper details the chatbot's technical customization to meet course-specific needs, provides evidence-based insights into practical challenges and opportunities, and offers strategic recommendations for effective AI-assisted pedagogy. Directions for further research are also outlined to explore and refine the role of generative AI in classroom settings. By examining two distinct courses, this study demonstrates how generative AI can be adapted across academic disciplines for more nuanced applications in education.

This paper examines a novel data-driven framework for spectrum monitoring in the context of non-public networks. To this end, we model an indoor scenario with a single non-cooperative transmitter (TX) and a spatially distributed network of sensor units (SUs) capable of recording the broadcasted signal as I/Q samples. Then, a dual-task deep neural network (DNN) is designed to jointly classify wireless signal waveform and estimate the TX coordinates, using raw data from the SUs as input. The numerical simulations focus on the model's scalability, generalizability, and robustness. In this context, we study the effect of the number of SUs and varying signal-to-noise ratio (SNR) on the accuracy of the developed DNN-based joint signal classification and TX localization algorithm. In particular, we show that an increased number of SUs improves the localization accuracy, whereas the performance of waveform classification remains nearly unchanged. Finally, we assess the model's complexity regarding the number of learnable parameters compared to single-task DNNs, demonstrating that extra functionality can be obtained at a low additional complexity cost.

The administration of 401(k) plans presents significant challenges due to the intricate regulatory frameworks established by the Employee Retirement Income Security Act (ERISA) and the Internal Revenue Service (IRS). These regulations necessitate precise compliance in areas such as contributions, distributions, and loan processing to avoid penalties and legal repercussions [1, 2].Traditional compliance systems rely on rigid rule-based mechanisms that often fail to address complex and evolving regulatory requirements. These systems are prone to inefficiencies, including high false-positive rates, which lead to increased manual effort and operational costs. Furthermore, they lack the adaptability to identify nuanced patterns of non-compliance, leaving organizations vulnerable to errors and oversight.Artificial Intelligence (AI) and Multi-Agent Systems (MAS) emerge as innovative solutions to these challenges. AI, particularly through machine learning (ML), offers advanced capabilities in anomaly detection, enabling organizations to identify potential compliance violations with greater accuracy. MAS, on the other hand, provides a decentralized approach to problem-solving, allowing multiple agents to coordinate and enforce compliance rules effectively [3]. Together, these technologies have the potential to revolutionize compliance monitoring by reducing manual intervention, improving precision, and ensuring audit readiness.This paper examines how AI and MAS can be leveraged to automate 401(k) compliance processes, emphasizing their ability to streamline operations and enhance regulatory adherence. By addressing the limitations of existing systems, it proposes a comprehensive framework for AI-driven compliance automation.

This paper presents a case study of a world-class national Information and Communications Technology (ICT) Incubator established in Saudi Arabia. ICT incubators, as the most rapidly growing segment of technology incubators, provide a cost-effective means for developing countries to jump-start their knowledge economies. As part of Saudi Vision 2030, initiatives are underway to transform Saudi Arabia into a knowledge-based economy and society. The BADIR-ICT incubator, established by King Abdulaziz City for Science and Technology (KACST) in collaboration with Saudi Telecom Company (STC), exemplifies the integration of digital transformation with national development goals. By offering a three-tiered service model, BADIR-ICT supports start-ups and small and medium-sized enterprises (SMEs) in overcoming challenges such as risk aversion, limited entrepreneurship culture, and restricted market access. Its incubation processes and business models, built on international best practices and tailored to the local context, align with Vision 2030's objectives of technological competitiveness, employment generation, and economic diversification. Key performance indicators (KPIs) are employed to evaluate its impact underscoring its strategic importance to Saudi Arabia's digital and economic transformation.

A document by Mohammad Zavid Parvez. Click on the document to view its contents.

The changing and rising uncertainty within power systems may result in more frequent and severe misalignments between hydropower storage management and society's interests. This work examines how risk aversion can influence the use of stored hydropower energy in centralized and liberalized power systems. We investigate the impact of different uncertainty sources and compare an equilibrium with risk-averse hydropower producers to risk-neutral and risk-averse central dispatch optimization approaches. Our main finding is that risk aversion can impact the scheduling strategies of competing hydropower producers in liberalized markets and those of a risk-averse central planner in opposite ways. Notably, risk-averse competing producers may have an incentive to spare less water in the cases of an expected energy deficit at the system level and higher price variability, potentially jeopardizing energy security. Meanwhile, widespread risk aversion in hydropower storage management can decrease or increase the producers' revenues at an aggregated level, depending on the energy situation.

The shift towards decentralized, power-electronicdominated renewable energy systems is leading to grids with lower inertia and reduced short-circuit ratios (SCR), creating new challenges for stability and control. Traditional simulation tools often operate in silos, focusing on specific areas like power flow optimization, grid stability, or controller design. However, the increased complexity of modern grids requires integrated co-simulation to assess the performance of controllers across larger, dynamic systems. This paper introduces a generic cosimulation framework using the OPC-UA protocol to integrate DIgSILENT PowerFactory with Python/Matlab, which is ideal for adaptive control design. The setup process is detailed, and two examples-classic rule-based and Deep Learning based adaptive controls-demonstrate its effectiveness. The entire pipeline is publicly available via GitHub.

Electromagnetics, radio frequency (RF), microwave, and antenna engineering are generally viewed as highly theoretical subjects, with a notable decline in student interest. This paper introduces a student-centred laboratory exercise for electromagnetic courses, developed as part of an IEEE Antennas & Propagation Society Educational Initiatives Programme. The proposed activity is based on transmission line analysis, measurements, and modelling. Students are required to model a set of 6 GHz transmission lines and passive networks using electromagnetic, closed-form, and their own equivalent circuit models. In addition, the modelling is complemented by measurements of varying quality and accuracy, allowing students to develop subject-specific critical analysis skills. Supporting instructors with instrument choice, we compare the measured response of the test lines using different inexpensive VNAs, evaluating their suitability to the different exercises. Thus, presenting for the first time guidance on which electromagnetic concepts could be illustrated using the different NanoVNAs. The exercises can be tailored for in-person or remote delivery. Implemented in a final-year undergraduate and Masters-level course, student feedback showed over 20% improvement in student satisfaction across all metrics surveyed.

In this work, the concept of test-time learning is presented, wherein Machine-Learning (ML) models are constructed by involving unlabeled test samples. Based on this concept, we propose an unsupervised method called Local Augment (LA) designed to improve the performance of trained outlier detectors at the prediction stage without altering the trained models or accessing the training data. LA operates under the only assumption that the model should produce similar outputs for similar inputs, implying that the prediction of a given sample can be enhanced by the predictions for its similar samples. Specifically, LA boosts outlier detection performance during prediction by fusing the outlier score of a given sample with the scores of synthetically neighboring samples generated by adding random perturbations to the given sample. This simple method demonstrates an average improvement of +0.04 Area Under the Receiver Operating Characteristic curve (AUROC) across 22 real-world datasets for all 11 tested detectors. Notably, this represents the pioneering work of enhancing ML models during the prediction stage without the need to modify the trained models or access the training dataset. This work opens up new possibilities for addressing existing bottleneck problems in various ML tasks beyond outlier detection in diverse domains.