The present work explores the geomorphotectonics of the Kathiawar (Saurashtra)-Kutchch Outlier in Gujarat, India north of the SONATA lineament. Using spatial statistics, the author explores intricate links between spatially resolved data, reveal patterns and correlations that drive the region's geological history. The work addresses an underlying issue, i.e., that of finding the spatial position of entities under study, which is a central issue in spatial analysis. The methodology combines several spatial analytic methods in order to derive a detailed representation of the geomorphotectonics of the area. Results of this study would make an important contribution to the ongoing task of understanding the geological history of the Kathiawar (Saurashtra)-Kutchch Outlier, providing insight to the complex processes that have created this peculiar geological unit.

The magnetic model of electric motors characterizes the magnetic flux as a non-linear function of the stator current as well as the rotor angle. An accurate magnetic model for electric motors is essential for high performance control strategies. The magnetic model is typically acquired by high-precision simulations or massive experiments of measuring magnetic flux data throughout the operating current range, and then applied in the control process via a look-up table of measurements. Both the acquisition and the application processes are time-consuming and may not be suitable for low-latency controls. To address this issue, we propose a novel compressed sensing-based method to acquire the magnetic flux map from limited randomly sampled data-points and further infer an analytical magnetic model. This analytical model can be used to efficiently compute the magnetic flux given any stator current instead of looking up measurement data in the control loop. The proposed approach is first validated on data simulated by finite element analysis (FEA) and then verified on data acquired on a permanent magnetic machine through experiments.

Contextual pathway reconfiguration introduces a dynamic approach to enhancing the performance and adaptability of transformer-based architectures in multimodal settings. By incorporating hierarchical gating mechanisms and auxiliary embedding layers, the proposed framework efficiently prioritizes and synthesizes information across diverse input modalities, addressing long-standing challenges in feature alignment and integration. Extensive experiments demonstrated consistent improvements in performance metrics across a variety of tasks, including image captioning and video question answering, showcasing the model's ability to generalize effectively while maintaining computational efficiency. The reinforcement learning-driven pathway evaluation mechanism emerged as a critical component, enabling real-time adjustments to computational pathways based on contextual input demands. Comparisons with traditional transformer methodologies highlighted the modular design's superiority in both scalability and robustness, particularly in handling noise-prone and unbalanced datasets. An in-depth analysis of latency and throughput further demonstrated the model's potential for deployment in high-demand, real-time applications. Error analysis revealed specific cases of improvement in cross-modal reasoning while identifying areas where input noise posed residual challenges. Ablation studies validated the significance of individual architectural components, confirming their contribution to overall model performance. The novel approach to dynamically reconfiguring attention mechanisms has broad implications for advancing multimodal applications, offering a versatile solution to integrate textual, visual, and auditory data. Additionally, energy efficiency assessments demonstrated the model's practicality for both training and inference scenarios, enhancing its viability for large-scale use. The findings present a compelling case for redefining multimodal transformer architectures to balance precision, efficiency, and adaptability, paving the way for more robust and versatile implementations.

The adoption of big data techniques in the capturing, processing and storage of health data was necessitated by the need to improve healthcare quality and patient outcomes.Big Data applications have huge potentials, especially with the availability of huge and growing healthcare data to transform and improve patient monitoring operations, which in turn can help in early detection of diseases, mitigation and the cure of diseases based on historic data. This could also help identify the best treatment for a patient’s specific circumstance, thus improving personalized treatment in patients.Designing effective Big Data applications in healthcare presents some challenges such as real-time data analysis, data privacy, data integration and system interoperability. This paper examines the challenges and opportunities associated with designing Big Data applications for healthcare monitoring, providing insights into the requirements for data architecture, interoperability, security, and real-time analytics.

The increasing integration of artificial intelligence (AI), learning analytics, and process mining has paved the way for enhancing the interpretability of educational data. This paper explores how these emerging technologies can synergize to improve the transparency and effectiveness of learning analytics, particularly in Massive Open Online Courses (MOOCs) and Free/Libre Open-Source Software (FLOSS) communities. A major challenge lies in the "blackbox" nature of AI models, which can hinder trust and adoption, especially in educational contexts. To address this, we delve into the role of explainable AI (XAI) in providing insights into learner behaviors while retaining the predictive power of complex machine learning algorithms. By reviewing state-of-the-art studies and synthesizing methodologies from process mining, AI, and learning analytics, this paper seeks to present a comprehensive framework for leveraging these technologies to make educational data more accessible and interpretable. The findings point toward future directions for combining process discovery techniques, explainability approaches, and educational data mining to create adaptive and transparent learning environments.

The objective of this study was to develop and evaluate an acyclovir-loaded ophthalmic lyophilisate carrier system, aimed at enhancing ocular drug delivery for the treatment of viral eye infections. Acyclovir, an antiviral agent, is often prescribed for ocular herpes simplex virus infections, but its ocular bioavailability is limited due to rapid drainage and poor permeability. To address these limitations, a novel lyophilized formulation was designed using biocompatible and biodegradable polymers as carriers to prolong the retention time and facilitate sustained release of the drug at the site of infection. The acyclovir-loaded lyophilisates were prepared using a freeze-drying technique, with optimization of the formulation parameters including polymer concentration, cryoprotectants, and lyophilization conditions. The characterization of the lyophilized systems included evaluation of physical properties, drug content, reconstitution behavior, morphology, and in vitro release profiles. The formulations exhibited high drug entrapment efficiency, desirable rehydration properties, and sustained drug release over an extended period. In vivo ocular tolerability of the developed system was assessed in a rabbit model. The formulations showed good ocular compatibility with minimal irritation or inflammation, indicating their potential for safe and effective use in ocular therapy. The pharmacokinetic study revealed enhanced corneal retention and prolonged drug release compared to conventional aqueous solutions. This study demonstrates the potential of acyclovir-loaded ophthalmic lyophilisate carrier systems for improved ocular drug delivery. The developed formulation provides a promising alternative to conventional therapies, offering prolonged ocular residence time, improved drug bioavailability, and minimal side effects, thereby enhancing the therapeutic efficacy of acyclovir in treating viral ocular infections. Further clinical studies are warranted to confirm these findings and explore the potential for broader clinical application.

The Internet of Bio-Nano Things is a promising technology to reduce turnover and increase resolution of medical data collection. Therefore, tiny devices, so-called Nano Machines (NMs) are placed inside the human body, where the collect, process and transmit information to devices outside the human body. Interfaces between in-body and the external electrical domain are crucial to the realization of this technology. In this paper, we propose sweat glands, a pre-existing infrastructure, as a novel interface through which NMs can transmit information from the inside to the outside of the human body. We provide a detailed mathematical model of the envisioned communication system and evaluate its communication performance using different types of detectors (e.g., neural network based detector).

Orthogonal frequency-division multiplexing (OFDM) has drawn a lot of attention in the study of integrated sensing and communications (ISAC). However, the fundamental drawback of Doppler intolerance limits its applicability in highly dynamic scenarios such as vehicular applications. In addition, the delay and Doppler shift that are non-integer times the resolution cause the off-grid effect, leading to scalloping loss and increasing the sidelobes. In this work, a super-resolution range-doppler map (RDM) modeling algorithm with low complexity is proposed to suppress these impacts. The proposed algorithm models the RDM contributed by the detected targets jointly via 2D surface fitting and then reconstructs and subtracts the corresponding signal components from the original received signal. Consequently, the weaker targets hiding under the inter-carrier interference (ICI) and off-grid effect become detectable. The communications capability is not impacted since the algorithm is only deployed at the radar receiver. In addition, two variants of the algorithm are illustrated, showing a trade-off between accuracy and complexity. Simulation results demonstrate that the proposed algorithms can effectively suppress ICI and mitigate off-grid effects, while simultaneously enhancing the accuracy of delay and Doppler estimation. It is indicated that the proposed algorithms can provide sufficient improvements in multi-target scenarios.

Autonomous rehabilitation support solutions, such as virtual coaches, should provide real-time feedback to improve motor functions and maintain patient engagement. However, fully annotated dataset collection for real-time exercise assessment is time-consuming and costly, posing a barrier to evaluating proposed methods. In this work, we present a novel framework that learns a frame-level classifier for real-time video assessment of compensatory motions in stroke rehabilitation exercises using weakly annotated videos. We consider three approaches: 1) a baseline approach that uses a source dataset to train a frame-level classifier, 2) a transfer learning approach that uses target dataset video-level labels and parameters learned from the source dataset with frame-level labels, and 3) a semi-supervised approach that leverages a target dataset video-level labels and a small set of frame-level labels. We intend to generalize to a weakly labeled target dataset with new exercises and patients. To validate the approach, we use two datasets with labels on compensatory motions: TULE, an existing video and frame-level labeled dataset of 15 post-stroke patients and three exercises, and SERE, a new dataset of 20 post-stroke patients and five exercises, created by the authors, with video-level labels and a small amount of frame-level labels. We show that a frame-level classifier trained on TULE does not generalize well on SERE (f1 = 72.87%), but our semi-supervised and transfer learning approaches achieve, respectively, f1 = 78.93% and f1 = 80.47%. Thus, the proposed approach can simplify the customization of virtual coaches to new patients and exercises with low data annotation efforts.  

As climate change drives global temperatures upward, seasonal sea ice in the Arctic is rapidly diminishing, forming meltponds. Meltponds absorb more solar radiation compared to snow due to lower reflectivity (albedo), accelerating the melting of the ice underneath. The manual assessment of the meltponds' complex boundaries is a demanding and timeconsuming endeavor. To streamline this process, we introduce a novel method for detecting and segmenting meltponds based on the UNet architecture, named ArcticNet. Our framework combines two UNets that incorporate recurrent, residual and attention operations. The first UNet extracts intermediate features that are further refined by the second UNet improving overall performance by learning hierarchical representations. The proposed architecture possesses the ability to enhance the boundaries of meltponds while accurately capturing their precise locations. The inclusion of residual and recurrent operations gives an expanded field of view for segmentation. Combining this with an attention gate, which concentrates primarily on target regions of diverse shapes and sizes, has been demonstrated to enhance the overall performance and robustness of the network by boosting model sensitivity and improving predictions. The performance of our model was assessed using high-resolution aerial images from Healy-Oden Trans Arctic Expedition (HOTRAX) and NASA's Operation IceBridge. The ArcticNet architecture categorized the HOTRAX and Operation IceBridge images into three and two classes respectively. HOTRAX include meltponds, open water, and snow, whereas Operation IceBridge include only meltponds and snow. The new algorithm showed superior performance in segmenting meltponds when compared to other state-of-the-art approaches.

The increasing adoption of edge computing in Internet of Things (IoT) and Cyber-Physical Systems (CPS) environments has made these systems attractive targets for ransomware attacks, as sensitive data is often processed at the edge. As IoT nodes are typically resource-constrained, traditional ransomware detection solutions are unsuitable for these environments. In this paper, we propose a novel lightweight ransomware detection framework for IoT edge devices using an optimized Deep Representation Learning Model (DeepLSE). Our model leverages advanced compression techniques, including quantization and pruning, to reduce the model size and power consumption, making it suitable for deployment on resourcelimited edge nodes. We utilize the CICAndMal 2017 dataset to train and evaluate our model, achieving an impressive 99.05% accuracy. Moreover, our model demonstrates significant advantages over conventional CNN and DNN approaches, with an inference time of only 6 milliseconds and a power consumption of 0.149 watts-demonstrating superior efficiency compared to other platforms, which consume significantly more power. The framework also includes a cloud-based device management system that handles edge node alerts and enables real-time control of infected nodes. Our research provides an efficient, real-time solution for ransomware detection at the IoT edge, ensuring both high accuracy and low resource usage.

This paper focuses on developing a novel machine-learning (ML) solution on edge-devices, operating in high-noise environments, to detect major faults and events (such as crashes). Every year, thousands of people are subject to bicycle crashes that result from motor-vehicle accidents, poor conditions, etc. A more reliable and on-site anomaly detection will allow for quicker responses in emergencies, helping reduce long-term injuries. Currently, statistical analysis models are being used, but they struggle to detect such anomalies in noisy environments, leading to false alarms. As a solution, an IoT edge device was developed that deploys an ultra-lite ML model to better detect crashes, then notify an emergency contact. To tackle the lack of open-source bicycle crash vibration data, this paper also proposes a novel bicycle crash simulation methodology, utilizing bicycle self-stabilizing properties, to get the data for training. The simple and cost-efficient crash-simulation methodology devised is shown to be rather effective, circumventing the project's cost and other safety-related limitations. The prototype consists of a microcontroller, an IMU, and a GPS. A separate device was also developed to do the data collection and data storage. An ultra-lite ML (<10 KB) was trained, tested, and developed in an iterative process. In a comparison with two statistical analysis models, the ML model had a 130% improvement in accuracy.

In this work, the Simplified Multiscale Model (SMSM) is incorporated into the Energy-Based (EB) vector hysteresis model to represent anhysteretic material behavior. This approach enables the inclusion of effects such as mechanical stress, magnetostriction, material anisotropy, and crystallographic texture. By integrating the anhysteretic model into the EB framework, it becomes possible to account for dissipative effects while utilizing detailed material information. To solve the EB model in conjunction with the SMSM, two approaches are pursued: a numerical optimization of a free energy functional and an explicit approximate variant, known as the Vector Play Model (VPM). Both methods are compared in terms of computational performance, and the differences in results are demonstrated through the simulation of the cross-section of an electric machine. Furthermore, the local as well as global behavior is investigated.

We present the reconfiguration of a circular feed horn operating in bandwidth from 1.4 GHz to 1.8 GHz for radio astronomical applications, adapting it to the 21 cm band (1.42 GHz), and performing observations of the hydrogen line. The proposed model was implemented using cylindrical structures. The design, simulation, measured performance, and reconfiguration are detailed. This feed horn is used at the prime focus of a parabolic reflector antenna. A prototype was designed and characterized, and the simulated and measured results were in good agreement. The reflection coefficients (S 11) of the feed horn was below-15 dB, and the voltage standing wave ratio (VSWR) remained below 1.43. The gains obtained from simulation and measurement were 12 dBi and 12.3 dBi, respectively. To determine the suitability of the proposed design in a practical environment, spectral measurements were performed using an FPGA-based spectrometer. The results showed that the intensity of the signal reached 3.5 dB more than that of the original feed horn. The overall performance of the proposed feed horn suggests that it can be applied to the optics of the 4.9 m reflector antenna.

Computed tomography (CT) is a widely used medical imaging modality which provides invaluable visual representation of various conditions ranging from neurological lesions such as haemorrhage, stroke, tumors etc. to cardiovascular disorders like calcium deposits, pulmonary embolism and many other pathologies. However, the ionizing radiation from the CT machine's x-ray tube has to be kept in check, because overexposure is related to elevated risks for genetic mutation or cancer development. In this work, we attempt to reduce the radiation exposure required for high-quality CT image formation by establishing rank sparsity in principal components' domain and developing a compressed sensing framework based on a novel nonlocal and nonlinear low-rank principal component analysis technique in image denoising, which will be subsequently incorporated as a building block for a sparse-view CT image reconstruction framework under the umbrella of convex analysis. Experiments will show that the proposed strategy provides a viable solution for low-dose CT, outperforming other wellknown nonlocal image restoration models in both denoising and reconstruction tasks. In particular, the proposed method will offer 4 − 10% improvement in root-mean-squared error relative to other nonlocal methods at little extra computational time.

The work reported in this article addresses the issue of reliability assessment from a systemic perspective. The general idea of the paper is based on the observation that failures occur at the level of elementary components but are induced by stresses emanating from the form of use of the system and the environment in which they are placed. Within this scope, a so-called U-Cycle approach is developed. It allows, in a matrix framework, to match the levels of decomposition of a system from a functional viewpoint and the resulting failures from a dysfunctional perspective. The modelling of this U-Cycle thus involves taking into account the repercussions of the systems' mission profiles in order to convert them into levels of stress on the basic components in a functional top-down analysis. Then, in a dysfunctional bottom-up study, it leads to characterise the impact of the failure of these components on the operation of the sub-systems to which they belong. The paper describes the principles of methodological instrumentation of this U-cycle based on the superposition of four descriptive views representing the structural, temporal, physical and logical forms of organisation. A case study based on the qualitative modelling of the reliability of an aircraft electrical flight control function is proposed to illustrate the developments.

Alzheimer's Disease (AD) is a prevalent neurodegenerative condition where early detection is vital. Handwriting, often affected early in AD, offers a non-invasive and costeffective way to capture subtle motor changes. State-of-the-art research on handwriting, mostly online, based AD detection has predominantly relied on manually extracted features, fed as input to shallow machine learning models. Some recent works have proposed deep learning (DL)-based models, either 1D-CNN or 2D-CNN architectures, with performance comparing favorably to handcrafted schemes. These approaches, however, overlook the intrinsic relationship between the 2D spatial patterns of handwriting strokes and their 1D dynamic characteristics, thus limiting their capacity to capture the multimodal nature of handwriting data. Moreover, the application of Transformer models remains basically unexplored. To address these limitations, we propose a novel approach for AD detection, consisting of a learnable multimodal hybrid attention model that integrates simultaneously 2D handwriting images with 1D dynamic handwriting signals. Our model leverages a gated mechanism to combine similarity and difference attention, blending the two modalities and learning robust features by incorporating information at different scales. Our model achieved state-of-theart performance on the DARWIN dataset, with an F1-score of 90.32% and accuracy of 90.91% in Task 8 ('L' writing), surpassing the previous best by 4.61% and 6.06% respectively.

In this contribution, we formalize the general case of linguistic summaries with multiple summarizers and qualifiers. We extend the definitions present in the literature to handle cases in the following form: Q R1 ⋆. .. ⋆ R k y's are P1 ⋄. .. ⋄ P l where ⋆, ⋄ ∈ {AND, OR}. Moreover, we study the consistency of fuzzy summaries, focusing mainly on the double negation property (DN). We consider different negations for linguistic quantifiers and show which can be used for preserving the DN property.