Jilin Provincial Academy of Forestry Science

Early and accurate diagnosis of melanoma remains a major challenge due to the heterogeneous nature of skin lesions and the limitations of traditional diagnostic tools. In this study, we introduce HyperFusion-Net, a novel hybrid deep learning architecture that synergistically integrates a Multi-Path Vision Transformer (MPViT) and an attention U-Net to simultaneously perform melanoma classification and lesion segmentation in dermoscopic images. Unlike conventional CNN-based methods, HyperFusion-Net combines the general feature extraction capabilities of transducers with the spatial accuracy of the U-Net, which is enhanced by a mutual attention fusion block that facilitates the effective fusion of semantic and spatial features. The model was trained and evaluated using four public ISIC datasets containing over 60,000 dermoscopic images. Preprocessing techniques such as hair removal, clipping, and normalization were applied to improve robustness. Experimental results show that HyperFusion-Net consistently outperforms state-of-the-art models including U-Net, DeepLabV3 + , TransUNet, and Swin-UNet, achieving superior performance in classification (accuracy: 93.24%, AUC: 95.80%) and segmentation (Dice coefficient: 0.945 in ISIC 2024). Ablation studies confirm the effectiveness of the multi-path design and fusion strategy in enhancing diagnostic performance while maintaining computational efficiency. Furthermore, the model demonstrates strong generalizability across datasets with different lesion types and imaging conditions.

To address low-efficiency trust management in the Internet of Vehicles (IoV), this paper proposes a hybrid consensus algorithm combining Delegated Proof of Stake (DPoS) and Practical Byzantine Fault Tolerance (PBFT). By selecting dual primary nodes through DPoS voting and trust values, it enhances consensus security. The three-stage consensus protocol is optimized to a fivestage one, reducing node communication and improving efficiency. Experiments show the algorithm increases throughput by <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$1.46 \times$</tex> on average and reduces consensus delay by <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$1.54 \times$</tex>.

Species interactions affect ecosystem productivity. Positive interactions (resource partitioning and facilitation) increase productivity while negative interactions (species interference) decrease productivity relative to the null expectations defined by monoculture yields. Effects of competitive interactions (resource competition) can be either positive or negative. Distinguishing effects of species interactions is therefore difficult, if not impossible, with current biodiversity experiments involving mixtures and full density monocultures. To partition changes in ecosystem productivity by effects of species interactions, we modify null expectations with competitive growth responses, i.e., proportional changes in individual size (biomass or volume) expected in mixture based on species differences in growth and competitive ability. We use partial density (species density in mixture) monocultures and the competitive exclusion principle to determine maximum competitive growth responses and full density monoculture yields to measure species ability to achieve maximum competitive growth responses in mixture. Deviations of observed yields from competitive expectations represent the effects of positive/negative species interactions, while the differences between competitive and null expectations reflect the effects of competitive interactions. We demonstrate the effectiveness of our competitive partitioning model in distinguishing effects of species interactions using both simulated and experimental species mixtures. Our competitive partitioning model enables meaningful assessments of species interactions at both species and community levels and helps disentangle underlying mechanisms of species interactions responsible for changes in ecosystem productivity and identify species mixtures that maximize positive effects.

Calcium is an essential macronutrient for plant growth and development, and there is an optimal calcium concentration for plant growth. Calcium ion concentration changes create “calcium signals” that regulate plant growth through perception, decoding, transduction, and response processes. However, the mechanisms by which calcium signaling regulates photosynthesis are still not fully understood. In this study, Quercus acutissima seedlings were used to investigate the inhibitory effects of different concentrations of the calcium channel blocker lanthanum chloride (LaCl3) on photosynthesis and the underlying mechanisms. The results show that increasing LaCl3 concentration significantly decreased photosynthetic parameters, photosynthetic pigment contents, and photosynthetic product accumulation. Long-term water use efficiency decreased with increasing LaCl3 concentration, while instantaneous water use efficiency initially increased and then decreased. Structural equation modeling analysis indicated that LaCl3 concentration was significantly positively correlated with leaf calcium concentration in Quercus acutissima seedlings, while it was significantly negatively correlated with stomatal conductance, carotenoids, and soluble sugar content. The study concludes that LaCl3 directly inhibits the photosynthetic physiological processes of Quercus acutissima seedlings by blocking calcium signaling, providing insights into the regulatory mechanisms of calcium signaling in plant photosynthesis and a theoretical basis for the cultivation and application of Quercus acutissima under varying environmental conditions.

Wind power prediction is critical for the efficient utilization of wind energy and the stable operation of power grids. However, existing prediction methods struggle to handle the complex characteristics of meteorological data. To address this challenge, this paper proposes an entropy-driven high-dimensional feature gain XGBoost model. First, traditional meteorological data are transformed into information entropy from an informational perspective, enabling the capture of randomness and nonlinear relationships within the data. Next, an entropy-driven high-dimensional (EDHD) classification feature selection method is designed to identify the relationships between covariates and response variables in the information entropy. Subsequently, a feature-gain XGBoost model is constructed to dynamically determine the optimal splitting features and thresholds, achieving the entropy-driven high-dimensional feature gain XGBoost model (EDHD-FG-XGBoost). Finally, experiments using real wind farm data from the northeastern region of China demonstrate the feasibility and effectiveness of the proposed method.