Search Results (2,696)

The deployment of object detection models on resource-constrained edge devices remains a substantial challenge, primarily because conventional static networks expend the same worst-case computational cost on every input, regardless of intrinsic difficulty. This paper proposes an input-adaptive dynamic neural network architecture for object detection in embedded environments. The present study investigates two orthogonal axes of input-adaptive inference for embedded object detection: The system demonstrates depth adaptivity through the implementation of Early Exit, and width adaptivity via group-wise Adaptive Routing. The proposed framework is constructed on a frozen Ultralytics YOLO26s backbone and incorporates two YOLO-style early-exit heads positioned at approximately 33% and 66% of the backbone depth. Furthermore, the framework incorporates two Straight-Through Gumbel-Softmax routers, which are appended after Layers 4 and 8 with group-wise hard gating. Both axes additionally accept an explicit external control signal that allows the host system to override the input-conditional policy at inference time. The dual-mode design facilitates the functionality of the trained checkpoint as either an input-adaptive policy, in which the depth and width are determined per sample from the input distribution, or an externally controlled policy. The experimental findings demonstrate two strongly asymmetric input-adaptive policies on a frozen YOLO26s backbone. The early-exit profile reduces the compute per sample from 12.739 to 10.532 GFLOPs—a 17.32% reduction according to our in-house Conv2d/Linear MAC-based GFLOPs estimator—while preserving baseline accuracy (mAP50 = 0.1545 vs. baseline = 0.1528; ΔmAP50 = +0.0017, within evaluation noise; mAP50–95 = −0.0033). Evaluating the router-only profile in the same validator pipeline with a sparsity penalty of γ = 0.05 results in a 12.3% decrease in logical GFLOPs (from 12.739 to 11.172), while maintaining an accuracy level that is at or above the PEFT baseline (mAP50 = 0.2324 and mAP50–95 = 0.1040). In our small-domain PEFT setup, training the dynamic-policy modules yields per-checkpoint mAP shifts in this magnitude. Therefore, we interpret the width-axis accuracy result as preservation of the baseline rather than an improvement. Our contribution on the width axis is reducing computing power while maintaining baseline accuracy. Importantly, the router profile’s logical GFLOP savings are not currently reflected in wall-clock latency under our dense-kernel PyTorch implementation. Achieving practical speedup requires sparse-kernel deployment, such as structured-sparse kernels, TensorRT, TVM, or Triton paths. We will address this in future deployment-level work. Our results indicate that the depth axis can yield genuine end-to-end speedup today, while the width axis offers deployment-pending compute reduction. Full article

The ubiquitin-26S proteasome system provides a key mechanism for regulating protein turnover in plants and contributes to the control of diverse developmental and stress-related processes. Within this system, Skp1-Cullin1-F-box (SCF) E3 ligases rely on F-box proteins to confer substrate specificity, enabling selective and dynamic regulation of target protein stability. The large size and structural diversity of the F-box protein family in plants suggest extensive functional specialization, although many members remain poorly characterized. Here, we review recent advances in the understanding of F-box protein function, with a focus on their roles in plant development, stress adaptation, and immunity. Specifically, this review integrates findings across development, abiotic stresses, and immunity to highlight shared and diverging regulatory nodes and critically assesses the strength of substrate evidence to distinguish bona fide from putative F-box targets. We highlight how F-box proteins modulate key regulatory pathways, including phytohormone signaling, reproductive development, root architecture, and secondary metabolism, as well as responses to abiotic and biotic stresses. Emerging evidence indicates that F-box-mediated proteolysis acts as an important layer of control linking environmental signals to downstream transcriptional and physiological outputs. A better understanding of F-box protein substrates and regulatory networks is important for dissecting plant adaptive mechanisms and may provide molecular targets for future crop improvement strategies. Full article

Adaptive urban traffic management requires timely recognition of changes not only in total traffic volume but also in turning-demand composition. This study develops a structural-entropy framework for identifying sustained changes in turning-demand structural complexity. It further calibrates candidate trigger criteria for structure-adaptive intersection management. The framework defines normalized entropy at the intersection and approach levels and computes adjacent-window entropy differences. Percentile-based thresholds and a two-window persistence constraint are then applied to identify sustained candidate-trigger variables. A case study was conducted using six months of 5 min movement-level turning-flow data from 13 consecutive signalized intersections on an urban arterial corridor, producing 654,346 valid adjacent-window pairs. In the case dataset, the 85th-percentile thresholds were 0.141 at the intersection level and 0.265 at the approach level. The two-window criterion identified 26,955 sustained structural-complexity-change events, whereas a three-window criterion reduced the count to 8434. At the single-window linkage scale, 60.5% of intersection-level high-change windows had synchronous approach-level support, and 47.0% had at least one approach-level leading signal in the previous window. The results indicate that the proposed framework provides an interpretable pre-identification signal for candidate control triggers, which should be combined with operational and safety constraints before implementation. Full article

Large reservoir construction generates vast drawdown zones characterized by novel hydrological regimes that impose unprecedented selective pressures. While annual plants serve as pioneer colonists during secondary succession in these ecosystems, the mechanisms allowing their seeds to persist through prolonged anti-seasonal flooding remain poorly understood. We investigated how seed germination responses to extreme submergence are influenced by dormancy traits and phylogenetic history. We conducted a field experiment on 44 common annual plant species in the Three Gorges Reservoir drawdown zone. Seeds were subjected to maximum submergence depths of 0 m (control), 5 m, 10 m, 15 m, and 20 m, along the reservoir’s hydrological gradient. Post-submergence germination percentages were measured and analyzed using linear and Bayesian phylogenetic mixed-effects models, with seed dormancy status, seed type, season, and species’ phylogenetic relationships as explanatory variables. Submergence significantly reduced overall seed germination (p < 0.001), but more than 75% of species retained germination capacity even after 20 m of submergence. Germination percentage distributions shifted from near-normal to bimodal with increasing depth. Although the regression of squared PIC values against phylogenetic branch lengths showed a significant relationship, phylogenetic signal for germination percentages was weak and non-significant across all depths (Pagel’s λ < 0.101, Blomberg’s K < 0.228, p > 0.05). Bayesian models revealed that dormancy season significantly interacted with submergence depth (Estimate = −1.41, 95% CrI [−2.16, −0.67]). Seeds dormant during autumn-winter maintained stable germination percentages across depths, while germination of spring-summer dormant seeds declined significantly with increasing depth. Our findings demonstrate that annual plant seeds possess widespread, species-specific tolerance to extreme submergence. This tolerance is primarily driven by environmental filtering rather than phylogenetic history. The seasonality of dormancy is a crucial adaptive mechanism, enabling seeds, particularly those dormant in autumn-winter, to withstand the harsh conditions of the Three Gorges Reservoir drawdown zone. This study provides a functional trait-based framework for selecting suitable species for the ecological restoration of reservoir drawdown zones globally. Full article

Background/Objectives: Migraine may be associated with structural changes in the brain, including the cerebellum and brainstem. Some of these changes reflect the brain’s plasticity in adapting to migraine-related alterations, but others may influence the severity of migraines and resistance to treatment. Some studies report changes in cortical thickness among migraine patients, and focal cortical dysplasia (FCD) has been considered a possible cause of these changes. We argued that FCD could contribute to the development of migraine and the severity of its symptoms. To date, there has been no consistent report of FCD occurring in migraine patients. Case: A 29-year-old woman presented with a history of at least 19 years of high-frequency episodic migraine without aura. She experienced motion sickness during childhood and adolescence. Her condition worsened last year, evolving into chronic migraine, which was partially controlled by medications such as amitriptyline and rizatriptan, leading to high-frequency episodic migraines. An MRI conducted in 2024 showed a small area of signal abnormality in the left occipital lobe, believed to represent cortical dysplasia. A follow-up MRI after three months showed no changes in this area. She is currently diagnosed with high-frequency episodic migraine and demonstrated severe migraine-related disability, with a MIDAS score of 25, and a severe impact on daily functioning, with a HIT-6 score of 65. Conclusions: The case involves a worsening migraine that was somewhat alleviated by a pharmacological intervention. FCD may contribute to brain hyperexcitability in this case and her motion-related problems during childhood and adolescence. FCD could also play a role in the increasing severity of her migraines and her partial resistance to medication. Full article

Myalgic Encephalomyelitis (ME) is a chronic, multisystem disease characterized by systemic metabolic dysfunction and post-exertional malaise (PEM). In this study, we investigated the dysregulation of irisin, an exercise-induced myokine, and its potential antagonism by thrombospondin-1 (TSP-1). In a cross-sectional study (92 ME patients vs. 44 sedentary healthy controls), plasma irisin and TSP-1 levels were measured at baseline and after a 90 min mechanical stress challenge applied to induce PEM. ME patients exhibited significantly lower baseline irisin (p < 0.05) and a blunted exertional response (p < 0.05). Paradoxically, baseline irisin was an independent predictor of fatigue severity (β = 0.728, p = 0.018), with moderate-to-severe patients showing elevated levels of both irisin and TSP-1 (p < 0.05), suggesting a compensatory but ineffective response. Functional cellular dielectric spectroscopy indicated that TSP-1 inhibits irisin signaling in a concentration-dependent manner. Irisin signaling was markedly reduced by both αvβ5 blockade and HSP90α inhibition in this experimental system, consistent with a diminished ability to counteract TSP-1. Collectively, these findings support a model in which dysregulation of the irisin–TSP-1 axis contributes to metabolic dysfunction in ME. Elevated circulating TSP-1 levels are associated with symptom severity and are linked to impaired irisin signaling in an HSP90α- and αvβ5-dependent context. This interaction is consistent with defective metabolic adaptation and highlights a potential therapeutic target that warrants further validation to restore energy homeostasis. Full article

In this Reply, we address the criticisms raised by Franzini, Valdenassi, and Chirumbolo concerning our study on the effects of ozonized saline solution (O3SS) on microglial polarization and endothelial responses in vitro. We clarify that the primary aim of the original work was mechanistic, relying on rigorously controlled cellular models that are universally recognized as essential preclinical tools in translational medicine. We reaffirm the validity of our experimental approach, including the preparation and characterization of O3SS based on empirically validated methodologies, direct ozone quantification, and standardized protocols consistent with the existing literature and clinical practice. Concerns regarding ozone chemistry, dose relevance, and hypochlorite formation are addressed through analytical validation, biological threshold considerations, and the use of certified assays. We further justify the choice of BV2 microglia and HUVEC cells as established and widely used models for investigating inflammatory and vascular pathways under reproducible conditions. Statistical analyses, gene expression interpretation, and the absence of comparative pharmacological agents are discussed in the context of the study’s focused objectives. Finally, we place our findings within the established framework of ozone as an indirect pro-oxidant that elicits adaptive redox signaling (“oxidative eustress”), emphasizing the translational relevance of in vitro systems for elucidating early mechanistic events. Overall, we maintain that our study provides a robust, balanced, and evidence-based contribution to the understanding of ozone-derived redox biology. Full article

To meet the requirement for high-precision dynamic positioning of fully actuated vessels under wave-frequency disturbances, and to achieve clock-synchronous triggering for system analysis, decision-making and reliable communication, this paper proposes a minimum-time simultaneous triggering (MTST) scheme based on a modified self-adaptive observer. Firstly, the concepts of result-dependent event (RDE) and conflict are introduced to describe the internal coupling characteristics of the system and the continuous actuation behavior under the superposition of triggering signals. Then, for the estimation of yaw perturbation, a self-adaptive parameter algorithm is employed in the modified observer, whose stability is subsequently proven. To reduce channel occupancy during the cooperative transmission of distributed triggering signals, a multi-port scheme is proposed, including RDE, and a corresponding controller is designed. Furthermore, to avoid the computational explosion phenomenon and estimate complex nonlinear unknown terms, the dynamic surface method and radial basis function neural network are used in filtering and function approximation, respectively. Finally, theoretical derivations show that the multi-port processing ensures the stability of all system nodes without the Zeno phenomenon. Meanwhile, the MTST scheme also maintains system stability while effectively eliminating both the Zeno phenomenon and signal conflict. Numerical simulation results reveal that compared with the multi-port event-triggering (MET) scheme, the MTST scheme achieves performance improvements of 9.76%, 0.37%, and 43.15% in tracking precision, energy efficiency, and control smoothness, respectively, which demonstrates its prominent advantages in event-triggered control systems. While improving positioning accuracy, the scheme exhibits a slight slowdown in heading-direction convergence and introduces a heavier communication load. These characteristics reflect a fundamental trade-off: the MTST scheme provides superior control performance at the cost of an increased triggering frequency and greater communication overhead. Full article

This is a narrative review that explores the development of non-implantable vestibular devices designed to address postural instability, particularly in aging populations and patients with vestibular hypofunction. It establishes that balance relies on complex sensory integration and that the functional decline of this system creates a significant medical need. Three principal technological strategies are examined: sensory substitution devices, galvanic vestibular stimulation (GVS), and immersive visual feedback systems. Sensory substitution devices, which convert balance data into auditory, tactile, or electrotactile cues, demonstrate significant promise. Examples like vibrotactile belts provide feedback that reduces postural sway, enhancing stability and patient confidence. Parallel to this, GVS—using electrical currents applied to the mastoids—emerges as a potent non-invasive method to modulate vestibular pathways, improving balance control and even inducing neuroplastic changes, especially with stochastic “noisy” signals. The most recently developed devices include augmented and virtual reality technologies that offer innovative visual feedback, creating enriched rehabilitation environments that accelerate recovery by promoting sensory reweighting and neural adaptation. This review concludes that while implantable prostheses are advancing, non-invasive devices offer versatile, affordable, and complementary solutions for balance restoration. The future success of non-invasive alternatives hinges on developing more sophisticated stimulation protocols that account for the complexity of natural movement and individual patient contexts, expanding therapeutic options for vestibular disorders. Full article

Deep ConvNets suffer from gradient signal degradation as network depth increases, limiting effective feature learning in complex architectures. ResNet addressed this through residual connections, but these fixed short circuits cannot adapt to varying input complexity or selectively emphasize task-relevant features across network hierarchies. This study introduces GradAttn, a variation of the residual approach in CNNs that replaces the fixed residual connections with attention-controlled gradient flow. By extracting multi-scale CNN features at different depths and regulating them through self-attention, GradAttn dynamically weights shallow texture features and deep semantic representations. For representational analysis, we evaluated three GradAttn variants across eight diverse datasets: from natural images and medical imaging to fashion recognition. The results demonstrate that GradAttn outperforms ResNet-18 on five of eight datasets, achieving up to $+11.07\%$ accuracy improvement on FashionMNIST while maintaining a comparable network size. Gradient flow analysis reveals that controlled instabilities, introduced by attention, often coincide with improved generalization, challenging the assumption that perfect stability is optimal. Furthermore, positional encoding’s effectiveness turned out to be dataset-dependent, with CNN hierarchies frequently encoding sufficient spatial structure. These findings render attention mechanisms as enablers of learnable gradient control, offering a new way for adaptive representation learning in deep neural architectures. Full article

To improve the adaptability of single-phase inverters under different load conditions, this paper proposes a sensor-reduced control strategy based on unknown input observer (UIO). Under the assumption that the load current is bounded with known upper and lower bounds, the algebraic relationship between the capacitor voltage and load current is constructed by designing an interval observer. Furthermore, based on this relationship, a UIO is designed to realize the online estimation of inductor current and load current. Compared to existing control methods, the proposed scheme requires only the output voltage signal for sensing, effectively ensuring stable operation of the inverter under different load conditions while reducing system costs and improving reliability. Finally, simulation results verify the feasibility of the proposed approach. Full article

To maintain homeostatic conditions and optimal function during stressors, mitochondria initiate retrograde signaling. The mitochondrial integrated stress response (ISR) and unfolded protein response (UPR^mt) are critical quality control mechanisms activated during instances of mitochondrial perturbations. Restoration of mitochondrial homeostasis is orchestrated by three transcription factors, ATF4, CHOP, and ATF5, which upregulate protective genes to counteract stress. As the health and function of skeletal muscle are heavily dependent on a highly adaptive mitochondrial network, defining how mitochondrial health is maintained across various conditions is essential. Although several studies demonstrate the importance of these responses following instances of stress, the signaling mechanisms required to initiate such pathways remain poorly characterized in skeletal muscle. This review examines how the mitochondrial ISR/UPR^mt and related transcription factors respond to organellar stress by emphasizing the molecular events that occur during exercise, aging and muscle disuse. By consolidating the literature, this work aims to highlight the current understanding of mitochondrial stress response signaling within skeletal muscle and thus emphasize areas for future research and potential therapeutic strategies during divergent metabolic conditions. Full article

This paper investigates the output consensus problem in heterogeneous multi-agent systems. To address the challenges of traditional analytical methods in handling unknown dynamics and disturbances, a control framework is proposed that combines known model structures with a data-driven adaptive mechanism. The framework uses a distributed internal model to compensate for system heterogeneity and incorporates an event-triggered mechanism to reduce communication burden. To improve transient tracking performance, a reinforcement learning strategy based on centralized training and decentralized execution is introduced to adaptively optimize local feedback gains. Simulation results show that the proposed method effectively bounds closed-loop signals, achieves relatively fast convergence, and demonstrates some robustness and communication efficiency under process noise. Full article

Conventional adaptive active noise control (ANC) techniques, such as filtered-x normalized least mean square (FxNLMS), frequently run into issues when the noise environment changes, leading to longer reaction times. Moreover, fixed-filter approaches may lose the essential phase information necessary for efficient noise cancellation. This paper introduces 2D-BiSpecNet, a novel, effectively delayless feedforward active noise control system that uses a deep learning co-processor to address these difficulties. The technique converts one-dimensional audio signals into 64 × 64 bispectrum matrices, which transform sounds into visual representations. Therefore, it focuses on nonlinear quadratic phase couplings (QPCs), which provide robust and amplitude-independent views of the noise structure. The system then applies a quick multilabel classifier to examine these representations and immediately generates a control filter via 15 parallel subcontrol filters. The paper specifies a 5 × 5 convolutional receptive field that had the maximum efficacy. Simulations with real acoustic data indicate that this configuration yields an average noise reduction of −14.48 dB for aircraft noise, outperforming the usual FxNLMS technique by nearly 6 dB. The technology conducts classification and filtering nearly seven times faster than adaptive approaches, thus reducing convergence delays and delivering a more reliable and low-latency solution for noise-canceling environments. Full article

Underground coal fires are persistent subsurface hazards threatening energy resources. UAV thermal infrared remote sensing provides high-resolution observations of surface thermal anomalies, but these signals may be spatially offset from underlying fire sources. An integrated framework was developed for subsurface temperature-field reconstruction and surface–subsurface correspondence and offset analysis. Surface thermal anomaly centers were extracted using statistical thresholding, adaptive kernel density estimation, and intensity-weighted centroids. Subsurface temperature fields were reconstructed using an MGSM-RBF model that combines multi-Gaussian fire-source representation with residual correction. The framework was applied to the Sandaoba coal fire area using UAV thermal infrared data and 370 borehole temperature measurements from 39 boreholes, covering depths of approximately 0–85 m. Reconstruction accuracy was evaluated using spatially buffered cross-validation and compared with eight baseline methods. MGSM–RBF achieved the best performance, with RMSE = 92.49 $°\mathrm{C}$, MAE = 61.26 $°\mathrm{C}$, and $R^2$ = 0.81. Two surface thermal anomaly centers and three subsurface fire sources were identified, with primary combustion concentrated at 30–55 m depths. Surface anomalies were not vertical projections of subsurface sources. The horizontal offsets were approximately one-fifth to one-third of burial depth, reflecting depth-dependent and multi-source-controlled surface thermal responses. These findings support UAV-based coal fire interpretation and fire-control planning. Full article