Search Results (96)

Early-stage flame and smoke in forest fire scenes are often small, weakly textured, and easily confused with complex backgrounds, while many accurate YOLO-based detectors are difficult to deploy on resource-constrained FPGA devices. This study proposes YOLO-FSD, a lightweight detector derived from YOLOv4-tiny for FPGA-oriented deployment. It integrates inverted residual and depthwise separable structures, a lightweight semantic enhancement (LSE) block at the deep feat2 feature, a lightweight P₄ detection head, and a shallow detail compensation branch. On the combined test set of the D-Fire and New Fire and Smoke datasets, YOLO-FSD achieves a mean average precision at an intersection-over-union threshold of 0.5 (mAP50) of 69.36%, with 3.951 M parameters and 1.520 G multiply-accumulate operations (MACs). Compared with YOLOv4-tiny, mAP50 increases by 2.76 percentage points, while parameters and MACs decrease by 32.76% and 55.53%, respectively. For deployment validation, batch normalization (BN) fusion and 16-bit integer (INT16) parameter conversion are followed by fixed-point forward inference on a Xilinx Zynq-7020 FPGA, with a software-side parameter-quantization sensitivity analysis used as an intermediate check. FPGA raw output evaluation achieves 69.14% mAP50, only 0.22 percentage points below the PyTorch 32-bit floating-point (FP32) model, demonstrating the feasibility of FPGA-side forward inference. Full article

Electroencephalography (EEG)-based emotion recognition is an important biosensing technique for affective brain-computer interfaces (BCIs), mental-state assessment, and physiological monitoring. Existing methods often rely on a single spectral descriptor or regular two-dimensional brain maps, which makes it difficult to jointly model local spatial representations, global spatial relations, and temporal dynamics. This paper proposes GLSTNet, a global-local spatial relations and temporal dynamics network for EEG emotion recognition. EEG trials are divided into short windows, from which multi-band spectral features are extracted and arranged into compact spatial maps. The local spatial encoder (LSE) learns local spatial and spatial–spectral representations from these compact multi-band spatial maps. The global spatial-relation encoder (GSRE) models long-range spatial relations between non-adjacent electrodes using a Pearson correlation prior and a learnable residual adjacency matrix. After local and global representations are integrated through gated fusion, the temporal dynamics encoder (TDE) models consecutive EEG windows using a gated recurrent unit with temporal attention. Comprehensive validation is conducted on two public EEG emotion datasets, the Database for Emotion Analysis using Physiological Signals (DEAP) and the SJTU Emotion EEG Dataset (SEED). In the subject-dependence setting, GLSTNet achieves 93.50 ± 3.22% accuracy for valence and 93.79 ± 3.64% accuracy for arousal on DEAP, and 92.48 ± 3.30% accuracy on SEED. In the subject-independence setting with target-subject calibration, GLSTNet obtains 75.61 ± 6.19% and 79.57 ± 5.99% accuracy for DEAP valence and arousal, respectively, and 88.22 ± 4.70% accuracy on SEED. These results indicate that integrating global-local spatial relations with temporal dynamics provides an effective representation strategy for EEG-based emotion recognition. Full article

Multi-expert large language model systems need to dynamically distribute user requests among multiple candidate experts to improve inference accuracy and system efficiency in complex task scenarios. To address the issues of existing routing methods relying on manual rules, unified discriminant boundaries, or large-scale annotated data, as well as high maintenance costs during expert expansion, this paper proposes an expert routing method, BGMM-LSE, based on positive sample semantic distribution modeling. The method constructs a positive sample set using only the requests historically successfully processed by each candidate expert, and maps the requests into dense semantic vectors through a pre-trained text feature extraction model; subsequently, a Bayesian Gaussian mixture model (BGMM) is independently trained for each expert to characterize the capability distribution of its successful requests in the semantic space. During online inference, the system encodes the target request into a feature vector, calculates the log-likelihood score combining the retained Gaussian component parameters and the smoothed covariance matrix of each expert, and uses Log-Sum-Exp to aggregate the probability contributions of different components. Finally, the request is routed to the expert model(s) with the highest score. Experiments were evaluated based on seven tasks: Math, GSM-Symbolic, HumanEval, Mbpp, MMLU, AIME1983–2025, and HellaSwag. The results show that this method achieves an average routing accuracy of 86.25%. The experimental results demonstrate that BGMM-LSE can provide stable request distribution capabilities for multi-expert large language model systems while maintaining interpretability and scalability. Full article

Music emotion recognition (MER) can be formulated as a multimodal optimization problem that predicts an emotion label from coupled audio and lyric sequences. Existing methods typically perform unimodal learning or coarse global fusion, which overlooks fine-grained temporal-token correspondences between musical dynamics and lyric semantics. We propose MELT (Music Emotion Learning with Temporal token-level fusion), an optimization-driven framework with four modules: a BERT-based lyrics semantic encoder (LSE), a segment temporal encoder (STE) that models audio-segment dependencies via a Transformer, a token-level temporal fusion (TTF) module with gated cross-attention, and an emotion mood head (EMH) for four-class prediction. Training is conducted end-to-end by jointly minimizing a supervised classification term and an auxiliary cross-modal contrastive alignment term, yielding a unified objective that improves both class separability and representation consistency. On the MoodyLyrics benchmark, MELT achieves 87.6% weighted F1 for four-class emotion recognition (angry, happy, relaxed, sad), outperforming unimodal baselines and representative early/late fusion strategies. Ablation results further verify that temporal encoding, gated token-level fusion, and joint optimization each contribute to the final performance. Full article

Semantic segmentation of land use and land cover (LULC) in arid regions remains challenging due to severe class imbalance, fragmented spatial distributions, and high spectral similarity among different land cover types. These characteristics often lead to an information bottleneck in deep segmentation networks and hinder the extraction of discriminative semantic representations. To address these issues, we propose SDS-Former, a lightweight semantic segmentation network specifically designed for remote sensing imagery in arid environments. SDS-Former incorporates an SSM-inspired Lightweight Semantic Enhancement (LSE) module to strengthen contextual modeling and alleviate the loss of discriminative information in deep features. To tackle scale variations, a Dynamic Selective Feature Fusion (DSFF) module is employed in the decoder to adaptively weight and fuse high-level semantics with low-level spatial details. Furthermore, a Feature Refinement Head (FRH) is introduced to enhance boundary localization and improve the recognition of small-scale and sparsely distributed land cover objects. Extensive ablation and comparative experiments demonstrate that SDS-Former consistently outperforms representative semantic segmentation methods across multiple evaluation metrics. On the Tarim Basin dataset, the proposed network achieves a mean Intersection over Union (mIoU) of 82.51% and an F1 score of 86.47%, indicating its superior effectiveness and robustness. Qualitative results further verify that SDS-Former exhibits clear advantages in distinguishing spectrally similar land cover types and preserving the spatial continuity of ground objects in complex arid-region scenes. Full article

This paper presents a systematic asymptotic analysis of the least squares estimator (LSE) for the drift parameter in a fractional stochastic heat equation driven by fractional Brownian motion. Fractional Brownian motion, capable of capturing stylized features in financial markets such as long memory, has become an important modeling tool in financial econometrics and risk management. Based on continuous-time observations of the Fourier coefficients of the solution, we first establish the strong consistency and asymptotic normality of the estimator. We then construct an alternative estimator based on the LSE and analyze its asymptotic behavior. This study provides new asymptotic inference tools for stochastic systems with long-memory properties and extends the theoretical framework for parameter estimation in fractional stochastic partial differential equations. Full article

Lysine carboxymethyl cysteinate (LCC) has been identified as a glutathione (GSH) precursor for the use of cosmetic products, providing a defense against oxidative stress by elevating GSH levels, and mitigating UVB-induced pigmentation and barrier disruption. In this study, the protective efficacy of LCC on epidermal barrier integrity under UVB irradiation was systematically evaluated and its underlying mechanisms were investigated. Results from the UVB-exposed 3D living skin equivalent model (LSE) indicated that LCC effectively restored UVB-induced reductions in epidermal living cell thickness by 9.67%. In addition, LCC markedly increased the expression of key biomarkers related to cornified envelope (CE) formation and skin hydration, including transglutaminase 1, involucrin, loricrin and aquaporin 3 by 104.80%, 121.67%, 218.63% and 388.39%, respectively, compared with the UVB group. Transcriptomics analysis in human primary keratinocytes further revealed that LCC regulated multiple biological functions, including glutathione synthesis pathway, oxidation response, inflammatory process, and notably autophagy. After confirming LCC’s potential in boosting autophagy-associated gene expression (p-value < 0.05) and autophagy activity (p-value < 0.01) in keratinocytes, functional validation in the same model confirmed that LCC counteracted UVB-induced suppression of genes involved in barrier formation, particularly those associated with CE development and autophagy, while these protective effects were abolished by chloroquine, an autophagy inhibitor. Findings from the UVB-exposed LSE model further substantiated this mechanism. Collectively, these results demonstrate that LCC safeguards the epidermis from UVB-induced cornification abnormalities through the activation of autophagy. Full article

It might be difficult in many countries to find extended time series of measurements related to parameters of lakes’ hydrology and their interactions with catchments. Nowadays, the combined use of satellite imagery and spatially distributed hydrological models may contribute substantially to this direction. In this study, in order to assess for a long period of years a lake’s surface elevation (LSE) and its water balance components, Lake Kastoria and its catchment, under Greece’s dry-thermal conditions, were selected as the case study. This research employed the MIKE SHE coupled with the MIKE HYDRO River (MHR) hydrological modeling system, fed with precipitation and leaf area index (LAI) data coming from a ground weather station, typical values of LAI for the specific area, and satellite products from NASA for the precipitation and from Copernicus Global Land Service for the LAI. In all cases where satellite data were used, the simulation of the long-term LSE was very satisfactory, with minor to medium changes to the inflow and outflow components of the water balance in both the catchment (from 0.32 to 7.36%) and the lake (from 1.47 to 11.3%). The above changes were also reflected in the runoff coefficients. In conclusion, the above satellite products can adequately be used for the prediction of the LSE. Furthermore, a plethora of quantified information in relation to the catchment’s water balance can be extracted and used in decision-making processes. Full article

African elephants maintain a delicate balance between high intestinal helminth numbers and symptoms of poor health. However, calves and weaners face survival threats from damage caused by helminth infections and infestations during periods of starvation. Using coprological evaluations, we compared helminth infection patterns and selected risk factors in elephants from Tsavo and the Laikipia-Samburu ecosystem (LSE). Eggs and adult worms were identified and classified into genera using morphological and morphometric characteristics. Overall prevalence rates were high at 95.6% and 98.5% for Tsavo and Laikipia-Samburu, respectively, and 97.1% in total. Using the chi-squared test of independence, nematode prevalence rates (97.1%) were significantly different from trematodes (32.6%) (χ2(1) = 248.84, p < 0.001). Prevalence rates did not differ across age classes in the two populations. Using the Generalized Linear Model, sex had an independent significant effect on the observed mean worm burdens (F(1) = 5.842, p = 0.016, CI 95%), but neither age nor location did. However, the interaction had a significant effect on mean eggs per gram (EPGs) observed between age and location (F(1) = 14.043, p < 0.0005, CI 95%) and between age and sex (F(1) = 4.858, p = 0.028, CI 95%). Some of the identified helminth species may potentially lead to life-threatening gut mucosa hemorrhaging and ulcerations, particularly during stressful conditions, although the animals were not clinically evaluated. Full article

This research utilizes a modified cellulose nanocrystal composite as an adsorbent to remove cadmium (II) through a column study. A fixed-bed column was used to remove cadmium (II) at room temperature using varying process factors, such as pH (4–8), bed height (3–9 cm), flow rate (3–7 mL/min), and concentration (10–20 mg/L). According to these findings, cadmium (II) breakthrough occurred more quickly at lower bed heights, higher flow rates, and higher cadmium (II) concentrations. The Thomas model is the most appropriate kinetic model. Deep learning models, such as the adaptive neuro-fuzzy inference model with two algorithms (backpropagation and least squares estimation), were effectively used to model the effectiveness of cadmium (II) removal in aqueous solutions via modified cellulose nanocrystals. To compare the model’s predicted results with experimental data, statistical approaches were employed, including calculating the coefficient of determination (R²) and mean square error (MSE). The ANFIS model used to predict cadmium (II) adsorption via modified cellulose nanocrystals had a strong correlation value of 0.997 for least squares estimation (LSE) and 0.999 for the gradient descent (backpropagation) method, indicating the effectiveness of the trained model in predicting the cadmium (II) adsorption process. Full article

The growing use of electrical and electronic appliances, coupled with shorter product lifespans, has accelerated the rise in waste electrical and electronic equipment (WEEE). Accurate forecasting is essential for addressing environmental challenges, conserving resources, and advancing the circular economy (CE). This research employs a Trigonometry-Based Discrete Grey Model (TBDGM(1,1)) that integrates the Jaya algorithm and Least Squares Estimation (LSE) for parameter estimation. By leveraging Jaya’s parameter-free robustness and LSE’s computational efficiency, the model enhances prediction accuracy for small-sample and nonlinear datasets. WEEE data from Washington State (WA) in the USA and Türkiye are utilized to validate the model, demonstrating cross-context adaptability. To evaluate performance, the model is benchmarked against five state-of-the-art discrete grey models. For the WA dataset, additional benchmarking against methods used in prior e-waste forecasting literature enables a dual-layer comparative analysis, which strengthens the validity and practical relevance of the approach. Across evaluations and multiple performance metrics, TBDGM(1,1) attains satisfactory and competitive prediction performance on the WA and Türkiye datasets relative to comparator models. Using TBDGM(1,1), Türkiye’s e-waste is forecast for 2021–2030, with the 2030 amount projected at approximately 489 kilotones. The findings provide valuable insights for policymakers and researchers, offering a standardized and reliable forecasting tool that supports CE-driven strategies in e-waste management. Full article

Convolutional neural networks (CNNs) and graph convolutional networks (GCNs) have been widely applied to hyperspectral image classification tasks, but both exhibit certain limitations. To address these issues, this paper proposes a multi-scale feature fusion architecture (MCGNet). Symmetry serves as the core design principle of MCGNet, where its parallel CNN-GCN branches and multi-scale fusion mechanism strike a balance between local spectral-spatial features and global graph structural dependencies, effectively reducing redundancy and enhancing generalization capabilities. The architecture comprises four modules: the Spectral Noise Suppression (SNS) module enhances the signal-to-noise ratio of spectral features; the Local Spectral Extraction (LSE) module employs deep separable convolutions to extract local spectral-spatial features; Superpixel-level Graph Convolution (SGC), performing graph convolution on superpixel graphs to precisely capture dependencies between object regions; Pixel-level Graph Convolution (PGC), constructed via adaptive sparse pixel graphs based on spectral and spatial similarity to accurately capture irregular boundaries and fine-grained non-local relationships between pixels. These modules form a symmetric, hierarchical feature learning pipeline integrated within a unified framework. Experiments on three public datasets—Indian Pine, Pavia University, and Salinas—demonstrate that MCGNet outperforms baseline methods in overall accuracy, average precision, and Kappa coefficient. This symmetric design not only enhances classification performance but also endows the model with strong theoretical interpretability and cross-dataset robustness, highlighting the significance of symmetry principles in hyperspectral image analysis. Full article

Accurate thickness recognition plays a vital role in safeguarding the structural reliability of critical assets. Pulse eddy current testing (PECT), as a non-destructive method that is both non-contact and insensitive to surface coatings, provides an efficient pathway for this purpose. Nevertheless, the complex, nonstationary, and nonlinear characteristics of PECT signals make it difficult for conventional models to jointly capture localized high-frequency patterns and long-range temporal dependencies, thereby constraining their prediction performance. To overcome these issues, we introduce a novel deep learning framework, multi-scale residual dilated convolution, and bidirectional long short-term memory with a squeeze-and-excitation mechanism (MRDC-BiLSE) for PECT time series analysis. The architecture integrates a multi-scale residual dilated convolution block. By combining dilated convolutions with residual connections at different scales, this block captures structural patterns across multiple temporal resolutions, leading to more comprehensive and discriminative feature extraction. Furthermore, to better exploit temporal dependencies, the BiLSTM-SE module combines bidirectional modeling with a squeeze-and-excitation mechanism, resulting in more discriminative feature representations. Experiments on experimental PECT datasets confirm that MRDC-BiLSE surpasses existing methods, showing applicability for real-world thickness recognition. Full article

Salinity extremes (SEs) play a crucial role in marine ecosystems, ocean circulation, and climate variability. Understanding their distribution and drivers is essential for predicting changes in ocean salinity under climate change, particularly in dynamic regions such as the Hawaiian Islands, where mesoscale eddies significantly modulate water mass properties. This study investigates the three-dimensional characteristics of SEs and their responses to mesoscale eddies using mooring observations and sea surface salinity data. We find that high salinity extremes (HSEs) generally occur more frequently than low salinity extremes (LSEs) in the study region, though LSEs exhibit greater duration and intensity. Mesoscale eddies modulate SEs significantly—anticyclonic eddies (AEs) enhance LSEs, whereas cyclonic eddies (CEs) promote HSEs in the upper layer. This relationship reverses in the deeper layer, with AEs favoring HSEs and CEs enhancing LSEs. These opposing effects are driven by a vertical displacement of the subsurface salinity maximum layer, where CEs lift high-salinity subsurface water to the upper ocean via upwelling, creating HSEs in the upper layer and LSEs in the deeper layer, while AEs subduct high-salinity water downward, reducing upper-layer salinity (LSEs) but increasing deeper-layer salinity (HSEs) via downwelling. Spatially, CEs exhibit a single-core high-salinity anomaly, displaced westward by 0.3 times of the eddy radius from the eddy center, with HSEs peaking in frequency and intensity near the core. In contrast, AEs display a dipole salinity anomaly (low northwest/high southeast), aligning with LSE frequency distribution, while HSEs show an inverse pattern. Mooring data further reveal that AE-LSE co-occurrence is highest within 1.2 times of the eddy radius, whereas CE-HSE probability declines with eddy intensity. Notably, AE-HSE and CE-LSE probabilities, though initially weaker, surpass AE-LSE and CE-HSE at certain depths, underlining the complexity of depth-dependent eddy modulation. These findings may advance understanding of ocean salinity dynamics and provide insights into how mesoscale processes modulate extreme events, with implications for marine biogeochemistry and climate modeling. Full article

This paper introduces a novel and flexible class of continuous probability distributions, termed the Alpha Power Survival Ratio-X (APSR-X) family. Unlike many existing transformation-based families, the APSR-X class integrates an alpha power transformation with a survival ratio structure, offering a new mechanism for enhancing shape flexibility while maintaining mathematical tractability. This construction enables fine control over both the tail behavior and the symmetry properties, distinguishing it from traditional alpha power or survival-based extensions. We focus on a key member of this family, the two-parameter Alpha Power Survival Ratio Exponential (APSR-Exp) distribution, deriving essential mathematical properties including moments, quantile functions and hazard rate structures. We estimate the model parameters using eight frequentist methods: the maximum likelihood (MLE), maximum product of spacings (MPSE), least squares (LSE), weighted least squares (WLSE), Anderson–Darling (ADE), right-tailed Anderson–Darling (RADE), Cramér–von Mises (CVME) and percentile (PCE) estimation. Through comprehensive Monte Carlo simulations, we evaluate the estimator performance using bias, mean squared error and mean relative error metrics. The proposed APSR-X framework uniquely enables preservation or controlled modification of the symmetry in probability density and hazard rate functions via its shape parameter. This capability is particularly valuable in reliability and survival analyses, where symmetric patterns represent balanced risk profiles while asymmetric shapes capture skewed failure behaviors. We demonstrate the practical utility of the APSR-Exp model through three real-world applications: economic (tax revenue durations), engineering (mechanical repair times) and medical (infection durations) datasets. In all cases, the proposed model achieves a superior fit over that of the conventional alternatives, supported by goodness-of-fit statistics and visual diagnostics. These findings establish the APSR-X family as a unique, symmetry-aware modeling framework for complex lifetime data. Full article