Rainfall prediction is a critical aspect of meteorology with significant implications for agriculture, water resource management, disaster preparedness, and climate science. This paper provides a comparative analysis of four primary rainfall prediction techniques: Numerical Weather Prediction (NWP) models, statistical models, machine learning models, and hybrid models. Each technique’s methodology, accuracy, strengths, and weaknesses are discussed in detail. The paper concludes by suggesting the most effective approach for researchers based on current advancements in data availability, computational power, and model sophistication.

My research optimizes three lightweight cryptographic algorithms—AES-256, ChaCha20, and Speck—for enhanced performance and security in IoT medical devices. By analyzing these algorithms throughput, latency, and memory usage on an Arduino board and applying specific optimization techniques, I have improved these algorithms and extrapolated the results to diagnostic, monitoring, and wearable devices. The core focus is finding the optimal tradeoff between security and performance for each optimized algorithm. The less memory usage the algorithm takes up in the device, the greater the performance. However, this also leads to lesser security. Finding the optimal balance will reduce computational overhead, improve processing speed, and lower energy consumption, leading to faster data processing. I strongly believe that these improvements to the algorithms will significantly enhance the functionality, security, and reliability of IoT medical devices, improving patient outcomes and advancing healthcare technology as a whole.

Determining the electric field distribution using electromagnetic solvers requires significant computational effort, a high time consumption, and operator training. In this paper, a deep learning (DL)-based method is specifically developed to reconstruct the electric field distribution generated by antennas, based on the dielectric properties of the breast. ResNet and UNet were employed to predict the electric field within complex breast structures. Nine distinct breast dielectric models were utilized to create the initial dataset. A novel data augmentation technique was applied and the size was increased by three to address the challenge of limited dataset size. The electric field distribution data were generated using commercially available finite element method-based electromagnetic simulations. DL architectures were customized to fit the study and trained over 4374 breast slices. ResNet demonstrated an average Signal-to-Noise Ratio (SNR) of 23.7±6.7 dB and an average Normalized Mean Squared Error (NMSE) of 0.0029±0.0039 across 504 test breast slices, while UNet averaged 24.55±6.64 dB and 0.0024±0.0031 respectively. Additionally, a masked loss calculation was incorporated to enhance the learning accuracy of the implemented models. This approach yielded an increase of 0.9 dB in average SNR for ResNet and a 0.8 dB SNR increase for the UNet architecture on average. These results indicate that the proposed DL models effectively reconstruct electric field distributions with substantial accuracy, highlighting its potential for practical applications in real-time microwave medical imaging and treatment planning. Further refinement and validation with a more diverse set of breast models could enhance the model's applicability and performance.

The rapid escalation of ransomware attacks has become a significant threat to individuals, organizations, and critical infrastructures, resulting in substantial financial losses and operational disruptions. A novel approach was developed through the application of machine learning techniques combined with file entropy analysis to provide an early detection system capable of identifying ransomware infections before critical files are encrypted. The method leverages entropy as a statistical measure of file randomness, allowing for the detection of encrypted or disordered files that are indicative of ransomware activity. Machine learning models, such as Random Forests and Support Vector Machines, were trained to classify files based on entropy and additional metadata, demonstrating high accuracy, precision, and recall across a diverse dataset of benign and ransomware-infected files. The results indicated that entropybased detection outperforms traditional signature-based methods, particularly in identifying zero-day ransomware variants. Future work includes the development of real-time monitoring systems for continuous entropy assessment, as well as exploring the use of deep learning to enhance detection performance. Integrating these advancements into practical, real-time environments holds significant promise for mitigating ransomware attacks before substantial damage occurs.

Gait analysis contributes to the identification of structural and functional limitations, as well as to the development of treatment plans and the evaluation of their effect on the patient. For this purpose, spatial and temporal characteristics of the patient's gait are studied and compared with reference values taken from control subjects. This paper examines the use of trajectories described by the instantaneous center of rotation (ICR) of the lower extremity segments as a biomechanical measure for assessing human gait. Gait trials were performed in which subjects walked on the ground and on a treadmill using a wearable inertial motion capture system (MVN Awinda, Xsens), while sports cameras recorded motion in the sagittal plane. Kinematic data from gait trials were acquired by video processing using an open source sports analysis tool (Kinovea) and directly through the Xsens Analyze software. The position of the ICRs of the lower extremity segments in the sagittal plane was calculated for each time sample and the determined trajectories were analyzed in time, frequency, and state space.

High-precision positioning in areas where Global Navigation Satellite Systems (GNSS) are degraded or unavailable is a necessity for the autonomous vehicles (AVs) of today and the near future and remains an active research problem. Fifthgeneration (5G) millimeter-wave (mmWave) technology presents a promising answer to wireless-based positioning in GNSS-denied environments. Like GNSS however, 5G positioning systems are expected to encounter brief signal outages in real, dynamic driving environments. During these outages, the positioning system must maintain its accuracy until a signal is available once more by relying on alternate technologies. On-board motion sensors (OBMS) including inertial measurement units (IMU)s and odometers are a logical solution to this problem, maintaining a position estimate through dead-reckoning methods. A classic solution is the integration of an odometer, or wheel encoder, with measurements from an IMU. Wheel encoders are limited by a fixed resolution and a relatively low data rate. Electronic Scanning Radar (ESR) are low-cost sensors found on most modern vehicles and measure the range, angle, and Doppler velocity of targets in their environment. In this paper, we explore the use of an ESR for forward velocity estimation as an alternative to the wheel encoder. ESR-based velocity estimation is integrated with 5G positioning, and its ability to maintain high positioning accuracy during 5G signal outages is assessed. Overall, due to an increased resolution and data rate, ESR velocity estimates were found to sustain a higher positioning accuracy during signal outages when compared to wheel-based odometry.

This paper proposes a bank of symbol-level correlators (SLCs) derived in accordance with joint maximum likelihood estimation (JMLE) to test for the presence of individual user equipments (UEs). A fully digital, extended, fractionally spaced, two-dimensional (2D) array of SLCs is addressed to estimate both residual frequency errors (RFEs) and residual time errors (RTEs) simultaneously for multiple active UEs. The proposed technique exploits multi-rate signal processing to decimate (i.e., downsample) the received signal for reducing subsequent computational complexity and to interpolate (i.e., upsample) in the frequency domain to enhance the accuracy of RFE estimation. Coherent accumulation of the squared magnitudes of SLC outputs further improves the signal-to-interference-plus-noise ratio (SINR). The proposed approach offers flexibility and practicality, enabling highly accurate estimation of RFEs and RTEs over extended ranges beyond one subcarrier spacing and beyond one symbol duration, respectively, while achieving precision at the fractional levels of a subcarrier spacing and a symbol duration. Simulation results confirm effectiveness of the proposed technique.

A document by ye yang. Click on the document to view its contents.

In this paper, we study the equilibrium problem of multi-mode traffic networks (auto and transit networks) with the tradable credit scheme(TCS). We propose a mathematical programming model with the Logit function as the mode-split function to study the users' travel mode and route choice behavior in the multi-mode network under the TCS. Based on the proposed network equilibrium model, Pareto-improving is further investigated. We find that the Pareto-improving under the TCS is a sufficient condition for reducing the system's total cost and we develop a Pareto-improving scheme for specific user categories. For the Pareto-improving scheme, all users do not become worse, and some or all users become better. In addition, a bi-level programming model is also proposed. The upper level minimizes the social total and transit operating costs by optimizing the number of free credits issued and the frequency of transit departure. The lower level is the traffic equilibrium assignment model of the multi-mode networks to determine traffic flows. The model is solved by combining the genetic algorithm(GA) and the heuristic algorithm based on the Method of Successive Averages (MSA). The correctness and validity of the models, propositions, and algorithms are verified using two numerical examples.

In this paper, we introduce an enhanced textual adversarial attack method, known as Saliency Attention and Semantic Similarity driven adversarial Perturbation (SASSP). The proposed scheme is designed to improve the effectiveness of contextual perturbations by integrating saliency, attention, and semantic similarity. Traditional adversarial attack methods often struggle to maintain semantic consistency and coherence while effectively deceiving target models. Our proposed approach addresses these challenges by incorporating a threepronged strategy for word selection and perturbation. First, we utilize a saliencybased word selection to prioritize words for modification based on their importance to the model's prediction. Second, attention mechanisms are employed to focus perturbations on contextually significant words, enhancing the attack's efficacy. Finally, an advanced semantic similarity-checking method is employed that includes embedding-based similarity and paraphrase detection. By leveraging models like Sentence-BERT for embedding similarity and fine-tuned paraphrase detection models from the Sentence Transformers library, the scheme ensures that the perturbed text remains contextually appropriate and semantically consistent with the original. Empirical evaluations demonstrate that SASSP generates adversarial examples that not only maintain high semantic fidelity but also effectively deceive state-of-the-art natural language processing models. Moreover, in comparison to the original scheme of contextual perturbation CLARE, SASSP has yielded a higher attack success rate and lower word perturbation rate

Transformer-based Large Language Models (LLMs) are progressively entering specialized areas, including the management of optical networks. These models have the ability to automate critical tasks, optimize network performance, and interact with operators, transforming how decisions are made in real-time network environments. This article explores the transformative role of LLMs in optical networks, addressing their application in network design, failure prediction, reconfiguration, and quality of transmission (QoT) estimation. Furthermore, the integration of LLMs in operator-driven decision-making processes is examined, highlighting how these models can provide intelligent recommendations that enhance operational efficiency. Key challenges such as real-time processing, hallucination risks, and domain-specific model adaptation are also discussed, alongside the deployment strategies for LLMs in cloud, edge, and hybrid network architectures. The article concludes by exploring future directions in LLM-assisted optical network management, with insights on human-in-the-loop systems and continuous learning.

In the context of paralleled Silicon Carbide (SiC) dies at low stray inductance, the current imbalance proves to be exceptionally sensitive to the static parameters of the devices. Even under conditions of full layout symmetry, significant current mismatch persists. This issue becomes prominent in emerging technologies utilizing standard cell design and die embedding, characterized by low stray power loop inductance. Traditional packages with relatively high stray inductance do not exhibit this particular problem. Addressing the challenges posed by these low-inductance technologies, this paper introduces a passive current-sharing method employing AC-coupled inductor. The proposed method is demonstrated through the implementation of a highly integrated converter with power loop inductance less than 1 nH. Notably, the presented approach is straightforward to implement and achieves near-perfect current sharing among paralleled dies, thereby facilitating the realization of high-current rating converters.

Meta-learning, or "learning to learn", enables machines to acquire general priors with minimal supervision and rapidly adapt to new tasks. Unlike traditional AI methods that approach each task from scratch using a fixed learning algorithm, meta-learning refines the learning algorithm itself through experience across various tasks, enhancing transferability and generalization. This is especially valuable when data collection is difficult or costly, allowing for effective learning from task sequences while reducing the dependency on extensive target domain data. Consequently, meta-learning has emerged as a promising field in machine learning. Although existing surveys provide valuable insights into meta-learning, they often present methods and applications in isolation and lack coverage of the latest advancements. Given the rapid growth of the field, a comprehensive survey is both necessary and challenging. Moreover, meta-learning algorithms often remain disconnected, with no unified framework to explain how they facilitate "learning to learn". This survey seeks to bridge that gap by systematizing meta-learning research, offering a thorough overview of strategies to enhance understanding. Additionally, the paper reviews over thirty representative meta-learning methods across models, tasks, and applications, analyzing their characteristics and challenges. To illustrate method performance, we evaluate more than fifteen models on six problems spanning thirteen scenarios, emphasizing the importance of selecting appropriate meta-learning approaches for practical applications.

Differential privacy (DP) has been widely used in communication systems, especially those using federated learning or distributed computing. DP comes in the data preparation stage before line coding and transmission. In contrast to the literature where differential privacy is mainly discussed from the point of view of data/computer science, in this paper we approach DP from a communications perspective. From this perspective, we show the contrast and opposition between the MAP detection problem in communications and the DP problem. In this paper, we consider two DP mechanisms; namely, the Gaussian Mechanism (GM) and the Laplacian Mechanism (LM). We explain why we have ϵ-DP if we use the LM, while we must have (ϵ, δ)-DP if we use the GM. Furthermore, we derive a new lower bound on the perturbation noise required for the GM to guarantee (ϵ, δ)-DP. Although no closed form is obtained for the new lower bound, a very simple one dimensional search algorithm can be used to achieve the lowest possible noise variance. Since the perturbation noise is known to negatively affect the performance of the data analysis (such as the convergence in federated learning), the new lower bound on the perturbation noise is expected to improve the performance over the classical GM. Moreover, we derive the perturbation noise required for both the LM and the GM in case of the adversary having auxiliary information in the form of the prior probabilities of the different databases We show that having auxiliary information is equivalent to reducing the tolerable privacy leakage, and hence it requires more perturbation noise. Finally, we analytically derive the border between the region where GM is better to use and the region where LM is better to use.

Mixed Integer Programming (MIP) is a foundational problem in combinatorial optimization, with wide-ranging applications in industries like logistics, scheduling, and resource allocation. Traditional methods for solving MIP, such as branch-and-bound and branch-and-cut, face scalability issues, particularly for large and complex problem instances. Recently, hybrid approaches combining machine learning (ML) with classical optimization methods have shown significant promise in improving solver efficiency and solution quality. In this paper, we review the integration of ML into MIP solvers, focusing on three key areas: branch-and-bound enhancements, cutting plane generation, and heuristic optimization for approximate solutions. ML models, including supervised learning, reinforcement learning, and graph neural networks, are used to inform critical decision-making processes within the solver, leading to faster convergence and better performance. We highlight how ML-enhanced solvers reduce computation time, improve scalability, and adapt dynamically to different problem structures. Finally, we discuss challenges such as generalization across diverse MIP instances and seamless integration with existing solvers. The future of MIP solving lies in further refining these hybrid methods, making ML-driven solvers a powerful tool for solving complex optimization problems.

Detecting and characterizing oil spills in arctic environments is challenging due to the presence of sea ice. Methods based on Ground-Penetrating Radars and acoustic sensors were used in the literature to detect oil in ice-covered waters; however, their implementation is costly, and their capabilities are limited in terms of oil-thickness estimation. To address this problem, we propose a new low-cost sensing system based on a planar capacitive sensor that can detect the presence of oil under ice sheets and measure its thickness and depth. Our proposed sensor is based on movable dual electrodes mounted next to each other on the same sensing plane. The electric field created between the electrodes extends beneath the sensing plane and is affected by the electric properties of the sensed material (ice, oil, or water). Changes in the mutual capacitance of the electrodes are related to changes in the thickness and depth of the embedded oil phase. The capacitance of the sensor is measured at two different excitation frequencies while changing the horizontal separation distance between the electrodes. These measurements are collected to create a dataset for training machine-learning-based classification and regression models to detect the presence of oil and measure its thickness and depth. In comparison with the available sensing techniques, our proposed sensor has several advantages, such as being noninvasive or non-intrusive, simple to manufacture, safe to operate, and having low cost and low maintenance requirements. The experimental evaluation described in this paper demonstrates the effectiveness of our proposed system, which showed a very high detection accuracy of more than 90% and an accurate thickness and depth estimation capability with a Mean Absolute Error (MAE) of around 0.5 cm for thickness and depth estimations for oil thicknesses ranging between 0.5 cm and 8 cm and for oil depths ranging between 2 and 5 cm.

Low-power and cost-effective IoT sensing nodes enable scalable monitoring of extensive, inaccessible, remote, and dangerous environments. However, some of these environments impose rough and extreme operating conditions on the nodes and require continuous adaptation and reconfiguration of physical and link layer parameters. In this paper, we closely investigate the stability of wireless links established between low-power IoT nodes deployed on the surface of different water bodies and propose dynamic transmission power adaptation to deal with link quality fluctuation and frequent disconnections. Our model is based on Minimum Mean Square Estimation (MMSE) and relies on statistics pertaining to received power and the acceleration the movement of water induces on communicating nodes. The model achieves a prediction accuracy exceeding 97%. The cost of processing and communicating the 3D acceleration statistics can be forgone with a reduced accuracy of 90% with a Kalman Filter.

This study presents a novel application of a Probabilistic Bayesian Neural Network (PBNN) for estimating vocal function variables and enhancing non-invasive ambulatory voice monitoring by addressing aleatoric and epistemic uncertainties in regression tasks. The proposed PBNN allows for estimating key physiological parameters including subglottal pressure, vocal fold contact pressure, thyroaritenoid and cricothyroid muscle activations, from seven aerodynamic and acoustic features. The PBNN is trained on the Triangular Body-Cover Model (TBCM) of the vocal folds to produce a non-linear inverse mapping between its inputs and outputs. Furthermore, the selected aerodynamic and acoustic features can be obtained in ambulatory settings, thus enhancing the practical applicability of the proposed method. Transfer Learning is then applied to integrate real voice data into the initially synthetic-trained network to refine subglottal pressure estimations. The confidence intervals generated by the PBNN illustrate its ability to identify uncertain estimations, as the results show correlations between prediction errors and the estimated aleatoric and epistemic uncertainties. This correlation is advantageous because it shows that the network can effectively predict potential inaccuracies in its estimations. Increased uncertainty is mainly observed at operating points where the TBCM is likely to exhibit non-linear behaviors, at higher subglottal pressures. This suggests that the selected input features may not be robust enough for capturing the nonlinear effects in the TBCM. These results highlight the potential for future research to assess the viability of incorporating new features and additional measurements that could better capture non-linear responses.