Search Results (1,956)

Deep hypergraph learning is evaluated almost entirely through leaderboards that rank methods by mean accuracy over a few random seeds, usually without significance testing. Is there a best hypergraph neural network, or does the apparent ordering reflect seed noise? We independently recomputed the node-classification track of DHG-Bench on a single GPU with twenty random seeds (against five upstream) and a different software stack, and applied a four-layer statistical audit to the per-seed accuracies: a reproducibility check, per-dataset paired Wilcoxon tests with Holm correction, an across-datasets Friedman/Iman–Davenport omnibus with Nemenyi and Holm-corrected pairwise tests, and a variance decomposition. Within a single dataset, twenty seeds distinguish most method pairs (74–98%), so the protocol is not underpowered. Across the nine datasets where all 17 methods complete, the omnibus rejects global equality (Kendall’s $W=0.45$), yet no pair survives Holm correction, and the top methods fall within one critical-difference band. One dataset carries more seed noise than between-method signal and cannot rank methods. The recompute also documents a non-reproducible method, a label-range data fault, and missing per-dataset configurations in the public release. No single method is statistically best across these datasets, so single-leader claims are not supported; we release a reusable significance-aware evaluation protocol. Full article

The bacterial proteome is a highly dynamic landscape rather than simply a static reflection of the genome. Recent research has revealed that proteome complexity extends far beyond canonical gene annotation, with N-terminal (Nt-)proteoforms emerging as an important underexplored additional regulatory layer. These molecular variants originate from a single genetic locus through alternative translation initiation at internal or external in-frame start sites, thereby generating N-terminal heterogeneity that can influence protein stability, subcellular localization, interaction networks, and the stoichiometric assembly of multiprotein complexes. While recent advances in riboproteogenomics, N-terminomics, and computational annotation strategies have enabled proteoform mapping at single-amino-acid resolution, rapid high-throughput discovery currently outpaces downstream functional characterization. This review discusses the technological advances driving Nt-proteoform discovery, including emerging ribosome profiling and proteogenomic approaches, and further evaluates strategies for the functional characterization of Nt-proteoforms. Particular emphasis is placed on the transition from conventional plasmid-based heterologous expression systems towards precise genome-engineering approaches that enable selective manipulation of alternative translation initiation events within their native genomic context. Such targeted strategies are essential to bridge the gap between Nt-proteoform identification and functional understanding, ultimately uncovering how individual bacterial genomic loci can encode proteoforms with distinct and potentially divergent functional roles in bacterial physiology and pathogenesis. Ultimately, we hypothesize that alternative translation initiation represents a biologically meaningful post-transcriptional regulatory mechanism that contributes to maximizing prokaryotic coding capacity without expanding genome size, rather than merely constituting stochastic translational noise. Full article

A double-ended traveling-wave accurate fault location method based on variational mode decomposition (VMD), generalized S-transform (GST), and a symmetric difference energy operator (SDEO) is proposed for single-phase-to-ground faults in small-current grounding distribution networks. The method is designed for fault conditions in which the fault current amplitude is low, the transient duration is short, and the initial traveling-wave wavefront is easily affected by high-frequency oscillation, reflection, refraction, and noise. The three-phase voltage traveling waves measured at both ends of the fault section are first transformed using Clarke modal transformation, and the voltage line-mode component is selected as the input signal. VMD is then used to decompose the nonstationary traveling-wave signal into several finite-bandwidth intrinsic mode functions. The high-frequency mode containing the initial wavefront mutation is processed using the generalized S-transform to enhance local time–frequency features. Finally, the SDEO instantaneous energy spectrum is used to calibrate the initial wavefront arrival time. For distance calculation, an improved double-ended traveling-wave location formula based on the horizontal section length and the absolute propagation time ratio at both line ends is constructed. This formulation reduces the dependence on a fixed empirical wave velocity and weakens the influence of line length deviation caused by practical line geometry. The method is verified using a deterministic PSCAD/EMTDC v5.0.2 and MATLAB R2021b co-simulation model with a fixed distribution network topology and arc-suppression-coil grounding. The tested cases cover selected fault distances, transition resistances, and fault inception angles. The simulation results show that the absolute location error remains within 100 m under all tested deterministic cases. The representative location error is 15 m at the 2.5 km fault point, 45 m under a 500 Ω transition resistance at the 1.5 km fault point, and 11 m under a 90° fault inception angle at the 6.15 km fault point. Compared with EMD-TEO, VMD-TEO, and VMD-GST, the proposed VMD-GST-SDEO method provides more stable wavefront calibration and lower location error in the studied cases. The results indicate the feasibility of the proposed approach within the stated simulation scope; further validation under stochastic noise, synchronization error perturbation, different sampling frequencies, and measured field waveforms is still required for engineering deployment. The fundamental novelty of the study lies in the task-oriented integration of VMD, GST, and the SDEO as a complete wavefront calibration chain and in coupling this chain with a propagation-time-ratio-based double-ended location formula for small-current grounding distribution networks, rather than in treating VMD, GST, or the SDEO as new standalone signal-processing algorithms. Full article

Illegal angling activities pose significant threats to aquatic ecosystem conservation and sustainable water resource management by disrupting ecological balance and aquatic resource protection, emphasizing the need for effective intelligent monitoring approaches. Illegal angling detection in complex aquatic environments remains challenging due to small target sizes, diverse human postures, and severe interference from shoreline vegetation, water reflections, and other complex backgrounds. To address these issues, this paper proposes an improved YOLOv8-based illegal angling detection framework, termed ERFA–YOLO. To enhance the discriminative representation capability of slender targets in complex scenes, an Enhanced Receptive Field Attention mechanism (ERFAConv) is introduced. By leveraging adaptive contextual perception and spatial geometric feature modeling, the proposed mechanism effectively enhances fishing-related target features while suppressing false activations from background noise. Furthermore, a temporal consistency-based post-processing strategy is introduced to reduce false positives caused by transient prediction noise and improve detection stability in dynamic aquatic environments. In addition, a dedicated illegal angling dataset covering multiple time periods, weather conditions, and complex shoreline environments is constructed to improve the generalization capability of the model in real-world natural scenarios. Experimental results demonstrate that, compared with the original YOLOv8 baseline, ERFA–YOLO achieves a Precision of 93.19% (+4.67%), a Recall of 86.24% (+1.34%), an ${\mathrm{mAP}}_{50}$ of 93.49% (+3.47%), and an ${\mathrm{mAP}}_{50:95}$ of 60.53%, while achieving real-time inference performance of 75.34 FPS on an NVIDIA RTX 4090 GPU. Compared with several mainstream object detection algorithms, the proposed method exhibits superior robustness and detection stability in complex natural environments, demonstrating the potential of ERFA–YOLO for intelligent illegal angling monitoring in sustainable aquatic resource management scenarios. Full article

Long-term implantable neural electrodes underpin brain–machine interfaces, deep brain stimulation, epilepsy monitoring, and closed-loop neuromodulation. Following chronic implantation, however, the foreign body response (FBR) at the electrode–tissue interface remains a major constraint on long-term performance, as reflected by increased interfacial impedance, lower signal-to-noise ratios, fewer resolvable units, and higher stimulation thresholds. This deterioration arises from interrelated events that include implantation injury, protein adsorption, blood–brain barrier disruption, complement activation, glial reactivity, oxidative stress, glial scar formation, and neuronal loss. It cannot be attributed solely to material ageing or encapsulation failure. This review examines the molecular mechanisms of neural-electrode FBR and relates them to surface-biofunctionalization strategies, including antifouling coatings, bioactive ligands, immobilized neurotrophic factors, drug-eluting electrodes, and emerging immunomodulatory interfaces. Establishing mechanistic links among molecular events, material interfaces, and functionalization strategies may guide the rational design of durable neural electrodes. Full article

This study proposes a surrogate-assisted framework to characterize system-level equivalent stiffness parameters of composite resilient floor subsystems from global resonance features, calibrated and assessed for consistency using actual building acoustic tests. Since intrinsic material characterization alone may not fully capture the contribution of resilient layers under complex in situ boundary conditions, a parametric finite element model utilizing a transversely isotropic formulation with five independent stiffness-related parameters is coupled with Gaussian Process surrogates and a genetic algorithm. Repeated laboratory impact tests provide empirical bending and torsion resonance targets near 60 Hz and 155–160 Hz. The trained surrogates predict these targets with a root mean square error (RMSE) below 0.02, enabling the inverse estimation to identify representative parameter sets within a ±5% tolerance. The optimization yields a finite admissible parameter region, reflecting the inherent non-uniqueness of the inverse problem. To establish framework consistency, the identified equivalent parameters are integrated into a hybrid vibro-acoustic scheme and assessed for consistency against sound pressure level measurements from real-world building tests. Crucially, the sensitivity analysis indicates a mechanical divergence: in-plane stiffness primarily governs resonance-frequency reproduction, whereas out-of-plane stiffness is more influential in the noise-related structural wall-response-energy assessment within the admissible solution region. This behavioral mismatch demonstrates resonance matching alone is insufficient for design prioritization. The proposed framework establishes a configuration-level tool to systematically evaluate structural improvement directions without direct intrinsic material testing. Full article

Cat cafés, where customers and resident cats share indoor spaces, create a distinctive yet understudied soundscape. This exploratory cross-sectional study examined cat cafés through three questions: acoustic conditions and customer perceptions using the ISO 12913 pleasant–eventful framework; associations between perceived sound sources, acoustic comfort, and mood; and links between perceptual ratings and typical activities. During weekday peak hours in July–August 2024, sound pressure levels were measured with NoiseCapture in three of four London cat cafés, while 50 customers completed questionnaires. SPLs generally ranged from 65–75 dB(A), exceeding the 45–50 dB(A) recommended for cafeterias; nevertheless, 88% of respondents described the soundscape as pleasant or very pleasant, mainly within the pleasant–uneventful region. Acoustic comfort was moderately positively associated with self-reported mood (r_s = 0.35, Holm-adjusted p = 0.01). Of six perceived source types, cat-generated sounds alone were positively associated with mood (r_s = 0.41, Holm-adjusted p = 0.02). Activity-related associations were non-significant after multiple-comparison correction, suggesting that larger samples may be needed. Attribute inter-correlations broadly reflected the established circumplex structure, preliminarily supporting the ISO framework in this setting. Overall, acoustic experience appears shaped not only by sound level but also by context, expectations, human–animal interaction, and possible biophilic associations. Full article

Conventional snapshot-type polarization devices often suffer from inherent ohmic losses caused by the metal structure, resulting in low utilization of system light energy. This research proposes a dual-layer sandwich architecture metasurface that integrates polarization control and high light transmittance for the mid-wave infrared 3~5 μm band. The top metal polarization-selective structures and the bottom dielectric hemispherical anti-reflection (AR) array are integrated monolithically on the same substrate. Specifically, the numerical simulations predict a peak transmittance of 95% at 4.4 and 4.8 μm, while maintaining extinction ratios ranging from 81.8 dB to 84.3 dB. This enables high extinction ratio polarization splitting while significantly broadening the transmittance flux of the device. It breaks the inherent trade-off between “high extinction ratio” and “high transmittance” in polarization devices, providing a high signal-to-noise ratio hardware foundation for high temporal resolution detection. Full article

Background/Objectives: Stroke-related aphasia significantly impairs communication and quality of life (QoL). Constraint-induced aphasia therapy (CIAT) and repetitive transcranial magnetic stimulation (rTMS) are evidence-based approaches targeting language recovery. No combined protocol exists for Greek, a morphologically complex language. The present study was designed as a feasibility pilot, with the primary objective of determining if a modified CIAT (mCIAT) protocol, with or without continuous theta burst stimulation (cTBS), could be delivered safely and acceptably in a Cypriot clinical context, while also reporting the first mCIAT protocol adapted for Greek. Methods: A single-subject experimental design (SSED) was employed. Two participants (PAs) with chronic anomic aphasia underwent a 10-day treatment protocol. PA1 received sham cTBS plus mCIAT-GR; PA2 received real cTBS (to the right inferior frontal gyrus) plus mCIAT. Multiple language, cognitive, and QoL measures were administered across three baselines and two follow-up time points. Results: Both PAs completed the full protocol with 100% session adherence and no adverse effects, establishing the feasibility of delivery. Neither participant showed improvements in standardized expressive language measures following mCIAT with or without cTBS; Standard Error of Measurement (SEM) analysis confirmed that observed score fluctuations on most measures fell within expected measurement noise. PA2 showed improvement in QoL (SAQOL-39g) domains at the two-year follow-up, though this finding requires cautious interpretation given the follow-up interval and regression-to-mean considerations. Conclusions: The Greek mCIAT protocol is feasible and safe to deliver in a Cypriot clinical setting. The absence of clear language improvements may reflect several factors, including ceiling effects, limited assessment sensitivity, insufficient treatment dosage, or limited intervention effects. These findings should be interpreted cautiously and require confirmation in larger, adequately powered studies. Full article

Conventional DC control and protection (C&P) systems rely on single-modal electrical data and are susceptible to false tripping and missed detection under complex operating conditions or novel fault types. In high-voltage direct current (HVDC) transmission, multi-modal data encompass time-series electrical quantities, unstructured transient waveforms, condition monitoring measurements, and environmental variables, each reflecting the system operating state from a distinct dimension with significant inter-modal complementarity. Nevertheless, fusing these heterogeneous modalities poses three key challenges: feature conflicts arising from data heterogeneity, difficulty embedding domain-specific C&P knowledge into data-driven models, and degraded model robustness under data noise and missing data conditions. This paper proposes a three-layer hierarchical fusion model that integrates multimodal data preprocessing, a C&P phase-aware rule-guided feature weighting strategy, and a dual-path decision mechanism. Experiments conducted on a dataset covering normal operation, typical fault, and complex operating scenarios demonstrate that the proposed model achieves an overall fault identification accuracy of 96.2%, which is 13.9 and 6.5 percentage points higher than a single-modal baseline and a generic multimodal model, respectively. The average decision latency of 7.2 ms satisfies the millisecond-level real-time requirement of industrial C&P systems, confirming the engineering applicability of the proposed approach. Full article

Smart wearable electrocardiogram (ECG) monitoring enables continuous cardiac assessment beyond clinical settings, but ECG signals are often degraded by baseline wander, muscle artifacts, and electrode motion artifacts. At present, pure end-to-end spiking neural networks (SNNs), with the advantage of low computational complexity, have rarely been explored for ECG noise suppression, particularly under dynamic conditions. To address this gap, this study proposes Direct-Coding-Physics-Aware Spiking Neural Network (DCPA-SNN) for wearable ECG denoising. The proposed method integrates a direct-coding SNN, channel attention, residual noise learning, and a physics-aware multi-domain loss function to preserve diagnostically important waveform structures. Clean ECG signals from the MIT-BIH Arrhythmia Database and real-noise segments from the MIT-BIH Noise Stress Test Database were used to construct single-noise and mixed-noise evaluation scenarios with input SNRs ranging from −6 dB to 4 dB to reflect the noise characteristics of wearable devices. Experimental results demonstrate that DCPA-SNN achieves robust denoising performance under different noise conditions. In the mixed-noise scenario, which serves as the primary evaluation setting of this study, the average denoised SNR reached 5.80 dB, with an average SNR improvement of 6.80 dB, while the R-peak detection rate increased from 90.71% to 95.72%. These results demonstrate that the proposed model, DCPA-SNN, provides a promising approach for wearable ECG denoising with potential for low-power deployment. Full article

Cloud removal is an indispensable preprocessing step in optical remote sensing. Reconstructing cloud-free imagery by combining multi-temporal optical observations with cloud-penetrating synthetic aperture radar (SAR) has become a mainstream approach. However, the existing studies mostly adopt simple composites, such as per-pixel least-cloudy selection or the temporal median, as baselines, and average accuracy metrics over entire scenes; together, these two practices may overstate the true gains of deep-learning methods. This paper proposes a temporal cross-modal cloud removal method (TCM-CR). In a multi-temporal sequence, the acquisition with the lowest cloud fraction retains true surface reflectance at its cloud-free pixels and is itself a high-accuracy baseline. TCM-CR exploits this baseline in two ways. First, on clear and light inputs, cloud-free pixels are taken unchanged from the reference image, so the true reflectance is preserved without loss, independent of training. Second, only cloud-covered pixels receive a bounded correction, in which SAR supplies the surface structure beneath clouds and multi-temporal observations are integrated along time while suppressing heavily clouded acquisitions. Experiments on the SEN12MS-CR-TS dataset show that TCM-CR maintains accuracy on par with the reference image on clear and light samples and improves the peak signal-to-noise ratio on heavy samples by 7.93 dB. In a cross-region experiment where one region is excluded from training entirely and used only for testing, heavy samples still improve by 7.27 dB. Full article

This paper investigates an active reconfigurable intelligent surface (RIS)-assisted integrated sensing and communication (ISAC) system with transmitter hardware impairments (HWIs) at the base station (BS). The active RIS provides both phase adjustment and amplitude amplification, which helps mitigate the multiplicative fading effect of passive RIS-assisted cascaded links and establish virtual line-of-sight (LoS) links for the target and multiuser communication users. Considering the coupling among the BS transmitter distortion noise, active RIS amplification noise, and the active RIS power constraint, we formulate a radar output signal-to-noise ratio (SNR) maximization problem. The radar output SNR is maximized by jointly designing the radar receive filter, the BS transmit beamforming matrix, and the active RIS reflection coefficients, while satisfying the quality-of-service (QoS) requirements of communication users, the BS transmit power constraint, and the active RIS power budget constraint. To solve the resulting non-convex problem with fractional objectives and high-order coupling terms, an alternating optimization (AO)-based iterative framework is developed. Specifically, the radar receive filter is updated using the generalized Rayleigh quotient, the BS transmit beamforming subproblem is handled by semidefinite relaxation and the Charnes–Cooper transformation, and the active RIS reflection coefficient subproblem is solved using the Dinkelbach transformation and majorization–minimization. Simulation results demonstrate stable convergence and show that the proposed design improves radar output SNR under BS transmitter HWIs. Full article

Illegal waste dumping remains a persistent urban management problem, yet enforcement is often reactive because cities lack spatially precise evidence on where risk is concentrated. This study presents an unsupervised machine learning and urban big data framework that converts routine administrative records into actionable patrol priorities. Using 1137 geocoded illegal-dumping complaint locations from Chuncheon, South Korea (January 2023 to May 2025), we integrate heterogeneous datasets describing residential intensity, population activity, educational facilities, waste disposal points, parks and public facilities, and monitoring infrastructure. A two-stage unsupervised pipeline is applied. First, DBSCAN aggregates nearby complaints to reduce spatial noise and supports feature screening by identifying effective influence ranges. Second, a grid-based K-means clustering classifies complaint occurrence areas into four interpretable typologies, selected using elbow and silhouette criteria. Hotspots are strongly associated with dense one-room housing, proximity to university districts, and high daily activity in mixed-use residential environments, while rural and predominantly non-residential zones show consistently low complaints. Infrastructure variables alone (e.g., disposal points, CCTV) have limited explanatory power. We derive a four-level enforcement priority scheme enabling targeted patrol and monitoring. Because complaint records reflect reporting behavior as well as underlying dumping activity, we interpret the resulting typologies as relative patrol-priority indicators rather than as a complete census of dumping risk. Full article

Personal health and safety are significant concerns across agricultural education, specifically agricultural mechanics. While hearing safety has been discussed for numerous years as a shortcoming in overall safety actions, it has been identified that attitudes and perceptions of sound and noise within agricultural settings are not aligned. This research continues work to focus on students’ understanding of and application of safety culture within an agricultural mechanics environment. Through direct and indirect exposure to noise levels, students reflected on and identified their use and intended use of hearing personal protective equipment (PPE). Pre- and post-instructional data aided in establishing students’ thresholds to use proper personal protective equipment and identify their perceptions of the decibel (dB) outputs for commonly used power tools in an agricultural laboratory setting. Course reflections alarmingly show a disconnect between intention and practice as many students do not utilize hearing PPE. While indicated threshold values and tool-specific PPE use align with the National Institute for Occupational Safety and Health’s recommended hearing PPE guidelines, the self-reported hearing PPE use results are concerning. While knowledge and application gaps still need to be addressed, based on this exploratory study, bounded in a single instructional context, continued exposure through direct and indirect instruction leads to increased knowledge of sound levels. Full article