Search Results (164)

In deep-space communications, conventional low-density parity check (LDPC) decoding algorithms suffer from slow convergence and high latency due to inefficient message scheduling. While residual-based layered belief propagation (RB-LBP) improves convergence, its serial scheduling limits hardware parallelism. Similarly, neural network-based decoders offer performance gains but impose prohibitive computational overheads on resource-constrained spaceborne platforms. To address these challenges, this paper proposes the parallel node-wise belief propagation (PNW-BP) algorithm. By employing a dynamic grouping strategy based on maximum residuals, PNW-BP constructs localized node sets for parallel message passing within sub-iterations. This strategy achieves intra-group parallelism while maintaining inter-group layered updates to preserve rapid convergence. Simulation results on the CCSDS standard (8176, 7154) LDPC code demonstrate that PNW-BP accelerates convergence, with no observable BER degradation under the evaluated conditions. For example, at $E_b/N_0=3.75\mathrm{dB}$ and a target BER of $5\times {10}^{-5}$, it requires only 8 iterations, whereas the NMSA, LBP, RB-LBP, and GS-BP require 37, 15, 16, and 20, respectively. Thus, PNW-BP provides an efficient solution for resource-constrained spaceborne applications. Full article

Accurate and low-cost fault diagnosis is essential for improving the reliability of cascaded neutral-point-clamped (NPC) inverters. This paper proposes a single-sensor fault diagnosis method for a single-phase three-module cascaded NPC inverter. Only one DC-side current sensor is required for the diagnostic algorithm, while the voltage sensor used in the outer voltage-control loop is not involved in fault-feature extraction. The measured DC-side current is decomposed via Fourier analysis, and a low-dimensional feature vector is constructed using the amplitudes of the zeroth, 2nd, 3rd, and 4th harmonics together with the phases of the 1st and 3rd harmonics. The six Fourier features are normalized using feature-wise Min–max parameters determined exclusively from the training data. A back-propagation (BP) neural network is then adopted to identify and locate 24 single-switch open-circuit faults in the three-module system. The investigated inverter produces 13 output-voltage levels under healthy operation, and the BP network converges after 5835 training iterations to an error threshold of 1 × 10⁻⁶. An adaptive confirmation criterion based on consecutive diagnosis-code consistency and inter-window feature convergence is introduced. For the nominal 25-class simulation test set, the accuracy, macro-precision, macro-recall, and macro-F1-score are all 100%. In addition, 134 of the 136 dynamic-condition simulation runs are correctly diagnosed, corresponding to an overall robustness-test accuracy of 98.53%. One confirmed, but incorrect final code occurs under the load disturbance applied at 90° of the output-voltage fundamental, and another occurs at an SNR of 20 dB, while no unconfirmed run is observed. Under the severe RL-load condition with τ/T₀ = 1, the mean and maximum diagnostic delays are 41.7 ms and 52 ms, respectively. Full article

Backpropagation is less a single algorithm than a pipeline of choices: how the error signal is propagated, how the weight gradient is assembled, and how the update is applied. This paper revisits three consecutive steps and proposes small, mathematically transparent modifications that improve gradient scaling and conditioning without changing the represented function class. The resulting method, Conditioned Backpropagation(CBP), combines (i) a layerwise gradient-norm equalization that counters the geometric depth dependence of the backpropagated error; (ii) an activation-centering reparameterization that removes the dominant rank-one mean term from the per-layer curvature; and (iii) a damped diagonal preconditioner that is positive-definite by construction. The composite operator is a bounded positive-definite preconditioner, so the method inherits standard nonconvex, Polyak–Łojasiewicz, and stochastic convergence guarantees at the per-step cost of ordinary backpropagation. No prior method composes these three repairs into one operator with a joint boundedness and positive-definiteness guarantee. Two further results, both new, concern equalization. On a block-structured strongly convex model, and for the curvature-equalizing target that the implemented gradient-energy equalizer approximates up to a quantified heterogeneity factor, equalization makes the convergence rate depth-uniform; the bounded-clip version that is actually run stays depth-uniform up to a clip-determined depth and retains a constant-factor improvement beyond it. Controlled experiments, run over ten or more seeds with paired significance tests, confirm the mechanisms: Equalization compresses an order-of-magnitude per-layer gradient disparity, centering cuts the top curvature eigenvalue about threefold and yields the lowest training loss, and the configurations combining centering with the damped preconditioner, including the full method, converge fastest. The effects persist on MNIST and on CIFAR-10 with a small residual convolutional network, at a measured per-iteration overhead below about twice that of Adam. Generalization is comparable across methods, and no end-to-end depth-scaling advantage is claimed, keeping the contribution focused on optimization geometry. Full article

Gradient-based inversion methods are widely used for time-domain airborne electromagnetic (AEM) data, but their results are commonly affected by the initial model and regularization-induced smoothing. These limitations are particularly evident when thin layers or alternating high- and low-resistivity structures need to be resolved. To address this problem, we propose a gradient-guided iterative enhancement (GGIE) inversion that combines a limited-iteration Gauss–Newton (GN) inversion with a lightweight U-Net. In each GGIE inversion iteration, the GN module first produces a coarse inverted model (IM) that preserves the main data-driven geoelectric trend but is still affected by regularization-induced smoothing. The trained U-Net then predicts a structurally enhanced model (PM) from the observed data and the IM. A data-misfit-guided adaptive approach is proposed to calculate the weight coefficients of the IM and PM and to construct an update model (UM). These coefficients are further smoothed by a momentum term so that the relative contributions of the IM and the PM are adjusted adaptively during the iterations. This design reduces error propagation from either component alone and dynamically balances learned structural enhancement with physics-based data consistency. The UM then serves as the initial model for the subsequent GN inversion. GGIE inversion is tested on synthetic data, and the results show that it is most beneficial for complex multilayer structures, for which it reduces the mean relative error and root mean squared error (RMSE) by 49.0% and 28.6%, respectively. Compared with the physics-informed neural network (PINN) baseline, GGIE inversion reduces the model relative error, log-domain RMSE, and data misfit by 19.4%, 7.0%, and 71.7%, respectively. Moreover, compared with U-Net alone, GGIE inversion reduces the data misfit by 85.4%. The proposed method is further applied to field data acquired from the Fox River area in Wisconsin, USA. The main advantage of GGIE inversion is its ability to resolve complex multilayered structures, thin layers, and sharp resistivity contrasts with improved accuracy and stability. Full article

This paper presents a novel method for pruning deep neural networks based on the concept of flow, derived from the continuous modeling of signal propagation across layers. We derive flow functions for fully connected, convolutional, and self-attention architectures, and we propose a new iterative pruning algorithm, Iterative Flow-Aware Pruning (IFAP), that leverages these measures to identify and eliminate non-essential parameters while preserving critical information pathways. Extensive experiments across ten prominent architectures (including CNNs, vision transformers, and efficient mobile networks) on ten benchmark datasets demonstrate consistent accuracy–compression trade-offs: $81\%$ of the evaluated configurations achieve a 60–$81\%$ reduction in computational cost relative to the corresponding baseline model. Furthermore, $97\%$ of the evaluated configurations retain more than $98\%$ of their baseline Top-1 accuracy. These results validate flow-based importance scoring as a robust and general-purpose foundation for model optimization. Full article

Deliberate coordinated cyber–physical attacks, which combine cyber intrusions with physical disruptions, pose growing risks to the secure and reliable operation of power systems. To identify highly disruptive coordinated cyber–physical attack strategies under pre-attack defense allocation and post-attack emergency dispatch responses, a tri-level defender–attacker–defender (DAD)-based attack strategy optimization model is proposed. First, based on the association between substation automation control systems and transmission line operation, the mechanism of breaker-tripping attacks (BTAs) through compromised digital relays in substations is investigated. Second, a tri-level DAD model is developed to optimize coordinated BTA and physical line attack strategies while accounting for pre-attack cyber and physical defense, cascading failure propagation, and post-attack emergency dispatch responses. Then, based on duality theory, network flow models, and the column-and-constraint generation (C&CG) algorithm, an iterative solution framework consisting of a defense master problem and an attack-dispatch subproblem is constructed to capture post-attack topology updates, cascading line outages, generator redispatch and load-shedding responses. Finally, the proposed method is validated using the IEEE 39 bus and IEEE 118 bus power systems. Case study results demonstrate that the proposed model identifies the most damaging coordinated cyber–physical attack strategies and that BTA-induced topology changes and cascading failure propagation significantly affect attack target selection. The proposed solution method obtains the accurate optimal objective value while reducing computational time by 92.91% for IEEE 39 bus power system, and it successfully obtains a converged solution for IEEE 118 bus power system. Full article

With advances in emerging material technologies, intelligent reflecting surface (IRS)-assisted vehicular networks have been gaining growing interest. By adaptively shaping the wireless propagation environment, IRSs can improve vehicular network quality of service (QoS). However, most IRS-assisted vehicular network studies are limited to individual RF or VLC frameworks, while only a few investigate IRS-assisted aggregated VLC-RF vehicular networks that combine wide RF coverage with high VLC data rates. In this paper, aggregated VLC-RF vehicular networks are supported by both optical IRSs (OIRSs) and RF IRSs, and a resource allocation scheme is developed to improve the total achievable rate. First, we establish a system model for IRS-assisted aggregated VLC-RF vehicular networks, and then formulate a problem to maximize the total achievable rate. Furthermore, we decompose the maximization of the total achievable rate into five subproblems and solve them iteratively via an efficient alternating optimization scheme based on block coordinate descent (BCD). Moreover, simulation results validate the convergence and efficiency of our algorithm, while highlighting the effects of crucial parameters on system performance, providing valuable insights for resource allocation in IRS-assisted aggregated VLC–RF vehicular networks. Full article

Reconfigurable intelligent surfaces (RIS) equipped with extremely large aperture arrays (ELAA) are emerging as a key technology for enhancing beamforming gain, spatial multiplexing, and angular resolution in sixth-generation (6G) wireless networks. When combined with non-orthogonal multiple access (NOMA), RIS can further improve spectral efficiency, system throughput, and energy efficiency. However, most existing studies on RIS-aided NOMA assume far-field propagation, where the incident wavefronts are approximately planar. In contrast, RIS-ELAA systems operating at millimeter wave (mmWave) experience spherical wavefronts in the radiative near-field regions. Also, it creates spatial non-stationarity and distance-dependent phase curvature. These effects invalidate the conventional monotonic path-loss assumption and make fairness-oriented NOMA design more challenging. This paper proposes a joint element and power optimization (JEPO) algorithm for near-field RIS-ELAA-assisted indoor NOMA systems, in which the RIS is dynamically partitioned into user-specific subarrays performing near-field phase synthesis toward the near user (NU) and far user (FU). The reversed far-to-near successive interference cancellation (SIC) ordering, governed by an effective FU channel gain greater than an effective NU channel gain, is formally established, and a closed-form optimal power allocation is derived by reducing the max-min fairness condition to a scalar quadratic in the target signal-to-interference-plus-noise ratio (SINR), eliminating iterative power search. Simulation results confirm that JEPO consistently outperforms four baseline schemes across transmit power, NU distance, angular separation, and RIS aperture size, with the largest gain observed at ${\theta }_{\mathrm{NU}}\approx {40}^{\circ }$ where fixed-partition baselines collapse to near-zero fairness while JEPO maintains robust performance. Full article

Underwater Wireless Sensor Networks (UWSNs) hold significant economic and military value; however, their routing protocols remain inherently vulnerable to external attacks. Unlike terrestrial networks, UWSNs cannot readily adopt complex cryptographic verification systems due to the high propagation delay, limited bandwidth, and low connectivity inherent in underwater acoustic channels. To address the wormhole attack—one of the most critical threats to UWSN routing—this paper proposes an intelligent routing scheme (UWSN-IRS) that not only detects wormhole attacks effectively but also identifies the source nodes and eliminates the threat. The proposed scheme comprises four integrated modules: a self-adjusting routing mechanism, a wormhole attack detection mechanism, a wormhole node localization mechanism, and an anti-cheating mechanism. The self-adjusting routing mechanism optimizes node distribution and intelligently searches for the optimal forwarding path. Upon the occurrence of a wormhole attack, the detection mechanism employs an artificial neural network to identify the compromised links and outputs a set of suspected wormhole nodes. Subsequently, the localization mechanism determines the exact positions of these malicious nodes through ranging and iterative positioning. Finally, the anti-cheating mechanism isolates the detected attacking nodes and deploys substitute nodes to fill the resulting monitoring voids. The experimental results demonstrate that the UWSN-IRS exhibits superior performance in attack scenarios, enabling reliable wormhole detection, precise attacker localization, and sustained normal network communication. Full article

In 3D integrated circuits, a placement decision that looks acceptable from a 2D wirelength view can still create a local thermal or stress problem after stacking. This issue becomes more visible as the number of tiers and the density of vertical interconnects increase. We propose ThermIC, a placement framework that brings thermal and mechanical risk estimates into the placement loop rather than treating them only as post-layout checks. The novelty of ThermIC does not lie in treating graph neural networks, reinforcement learning, uncertainty-aware learning, or physics-informed regularization as individually new techniques. Instead, ThermIC contributes a placement-time coupling mechanism in which physically typed graph propagation, dense multi-constraint risk prediction, and action-level reinforcement learning feedback are jointly organized for stacked 3D-IC placement. ThermIC uses a heterogeneous graph encoder to carry thermal, stress, timing, and congestion information through the netlist; a constraint head to estimate local hotspot, stress-risk, timing-violation, and congestion probabilities; and a sequential placement policy trained with physics-informed penalties. We evaluate the method on ThermIC-Bench, a simulated corpus with more than 30,000 finite-element samples from 18 heterogeneous 3D-IC designs with 4–8 tiers. Because the present study does not include proprietary industrial circuits, silicon measurements, or a tape-out case, the experimental results are interpreted as simulation-based benchmark evidence rather than final industrial qualification. ThermIC connects the heat-kernel branch to the discretized heat-conduction equation and the stress-filter branch to linear thermo-elastic equilibrium, providing a mechanism-level basis for physical interpretability. The analysis distinguishes offline simulation/training cost from online deployment cost and reports complexity, runtime, and memory scaling for practical large-scale use. Under joint DRC, thermo-mechanical stress, and thermally coupled timing checks, ThermIC obtains an $82.1\%$ physical verification pass rate. The peak-temperature error is $3.1$ °C, the hotspot localization IoU is 0.89, and the number of placement-closure iterations is reduced by $3.7\times$ relative to the heuristic baseline. Together, these benchmark results indicate that early, differentiable multi-physics feedback can make 3D placement less dependent on late correction cycles. Full article

Accurate gear fault diagnosis under variable operating conditions remains challenging because vibration signals are affected by noise, speed-load variations, and condition-dependent feature shifts. To address these issues, this study proposes a gear fault diagnosis framework that integrates multidomain vibration feature fusion with a back-propagation neural network optimized by the sparrow search algorithm (SSA-BP). Vibration signals collected from a planetary gearbox fault-implantation platform were used to identify seven health states, including normal condition, sun gear pitting, sun gear fracture, sun gear wear, planetary gear pitting, planetary gear fracture, and planetary gear wear. For each signal segment, a 20-dimensional feature vector was constructed by combining nine time-domain features, three frequency-domain features, and eight wavelet packet energy features. SSA was employed to optimize the initial weights and biases of a double-hidden-layer BP neural network before supervised training. Experimental results show that the proposed feature fusion scheme achieved a classification accuracy of 98.30%, outperforming single-domain and pairwise feature combinations. In overall fault classification, SSA-BP obtained 98.26% accuracy, 98.26% macro-recall, 98.27% macro-precision, and 98.26% macro-F1. Moreover, SSA-BP reduced the convergence iterations from 826 to 312 compared with traditional BP and maintained 95.18% accuracy under high-speed and high-load conditions with scarce training samples. These results demonstrate that the proposed SSA-BP model provides improved convergence efficiency, diagnostic accuracy, and cross-condition robustness for intelligent gearbox condition monitoring. Full article

Traffic sign segmentation is a fundamental component of intelligent transportation systems and autonomous driving, where reliable pixel-level perception is required under challenging real-world conditions such as illumination variations, occlusion, scale diversity, and complex urban backgrounds. In this work, we propose Residual–Recurrent Kolmogorov–Arnold Network U-Net (R2KAN-U-Net), where “R2” denotes the integration of residual convolutional learning and recurrent KAN-based feature refinement. The proposed architecture combines residual U-Net feature extraction, multi-scale KAN fusion, and recurrent KAN refinement to improve pixel-level traffic sign segmentation under challenging road-scene conditions. The proposed framework integrates three complementary components: (1) residual convolutional blocks for stable feature propagation; (2) a multi-scale KAN fusion bottleneck for capturing contextual information at different receptive fields; and (3) recurrent KAN refinement modules for iterative enhancement of discriminative features. Unlike conventional convolutional architectures, the proposed KAN-based formulation replaces linear transformations with learnable univariate functions, enabling adaptive nonlinear feature modeling. We conduct extensive experiments on a custom dataset containing 9300 annotated urban traffic scene images, as well as on the ADE20K and Cityscapes benchmarks. On the custom dataset, the proposed R2KAN-U-Net achieved a Dice coefficient of 0.92 and an IoU score of 0.89, providing a strong accuracy–efficiency trade-off for traffic-sign foreground segmentation. It achieves competitive segmentation accuracy compared with recent CNN-, transformer-, and state-space-based segmentation models while using fewer parameters and lower computational cost. Additional low-light experiments demonstrate improved segmentation stability, with R2KAN-U-Net achieving the highest low-light Dice score of 0.88 and a competitive low-light IoU of 0.79. Furthermore, the proposed architecture maintains competitive computational efficiency with only 24 M parameters, 44.8 G FLOPs, and near-real-time inference at 13 ms per image. The experimental results demonstrate that integrating KAN-based function-space learning with residual and multi-scale feature refinement provides an effective and computationally efficient solution for robust traffic sign segmentation in complex driving environments. Full article

Fake news detection has become critical for safeguarding social media users and maintaining a reliable news ecosystem. However, existing methods rely mainly on context information and propagation structure and do not consider the news structure from framing theory. As a highly structured genre, news implies writing intention and organizational logic in its discourse frame, which provides vital clues for authenticity verification. In this paper, we propose structure-guided iterative fusion with a transformer network for fake news detection (SIFTNet), which contains four modules: a structural label generator, an information architecture representation module, a structure-enhanced representation module, and a structure-guided iterative fusion module. Guided by framing theory, SIFTNet captures the semantics at both the local sentence level and global structure level. Extensive experiments demonstrate that our model achieves state-of-the-art performance on both Chinese and English datasets, exhibiting superior effectiveness and robustness. These findings validate the efficacy of applying framing theory to improve fake information detection. Full article

Object detection in wide-field-of-view (wide-FoV) unmanned aerial vehicle (UAV) imagery for Internet of Things (IoT) remote sensing applications requires accurate recognition of tiny objects under severe background redundancy and extreme scale variation. As the field of view expands, conventional dense detectors tend to waste substantial computation on non-informative regions, while feature downsampling and static receptive fields often cause the dilution of foreground information and scale confusion. To address these issues, we propose MFRC-Det, a unified framework built upon two complementary principles: mask-guided feature routing and adaptive context modeling. Specifically, a Superpixel-Masking Generator (SP-Masker) is introduced to estimate an image-space soft foreground prior by comparing Simple Linear Iterative Clustering (SLIC) superpixel histograms with a peripheral background reference, propagating the resulting scores on a superpixel adjacency graph, and projecting the refined region-level scores back to a pixel-level routing mask. Guided by these priors, a Greedy-Cutter (G-Cutter) converts dense feature maps into compact, foreground-focused patches without repeated backbone evaluation on cropped image regions, thereby reducing redundant background computation while preserving local structural coherence. On top of the retained regions, an Adaptive Receptive-field Selection Network (ARSNet) aggregates multi-scale contextual responses from several learnable receptive-field candidate branches. ARSNet predicts spatial selection weights conditioned on the input features, allowing each location to emphasize a suitable receptive-field response for object representation. Experimental results on VisDrone-DET and UAVDT demonstrate that MFRC-Det achieves competitive detection accuracy with favorable computational efficiency. Specifically, MFRC-Det obtains 36.1% AP, 60.4% $A{P}_{50}$, and 38.5 FPS on VisDrone-DET and 21.3% AP, 36.8% $A{P}_{50}$, and 37.4 FPS on UAVDT. These results validate the effectiveness of mask-guided feature routing and adaptive context modeling for wide-FoV UAV object detection and suggest their potential value for computation-efficient aerial perception in IoT remote sensing applications. Full article

With the deepening integration of 5G-Advanced (5G-A) technology into smart manufacturing, the large-scale deployment of dynamic terminals—such as mobile robots and automated guided vehicles (AGVs)—within industrial private networks introduces complex, time-varying penetration and path losses. This significantly degrades the accuracy of conventional signal quality and capacity estimation methods, which were primarily designed for static terminal scenarios, thereby posing substantial challenges to coverage and deployment planning of industrial 5G access points, with downstream implications for power capacity dimensioning. To address this problem, this paper proposes a coverage-driven base station deployment optimization method formulated as a combinatorial optimization problem. The study constructs a signal strength assessment and network throughput calculation model tailored for dynamic industrial environments. This model captures the joint impact of terminal mobility and environmental obstacles on signal propagation, thereby enabling more reliable estimation of coverage performance and power consumption. Furthermore, by formulating the base station placement optimization as a combinatorial optimization problem, and by introducing mechanisms for equivalent transformation of the objective function and data preprocessing, the proposed method substantially reduces redundant computations during heuristic iterations. Simulation results verify that, compared with conventional static planning approaches, the proposed scheme enhances both the accuracy and computational efficiency of deployment planning while maintaining coverage quality. This work provides a theoretical foundation and a practical methodology for deploying reliable and energy-efficient industrial 5G-A private networks. Full article