Explore the vast range of titles available.

Graph Convolutional Networks (GCNs) are powerful deep learning methods for non-Euclidean structure data and achieve impressive performance in many fields. But most of the state-of-the-art GCN models are shallow structures with depths of no more than 3 to 4 layers, which greatly limits the ability of GCN models to extract high-level features of nodes. There are two main reasons for this: 1) Overlaying too many graph convolution layers will lead to the problem of over-smoothing. 2) Graph convolution is a kind of localized filter, which is easily affected by local properties. To solve the above problems, we first propose a novel general framework for graph neural networks called Non-local Message Passing (NLMP). Under this framework, very deep graph convolutional networks can be flexibly designed, and the over-smoothing phenomenon can be suppressed very effectively. Second, we propose a new spatial graph convolution layer to extract node multiscale high-level node features. Finally, we design an end-to-end Deep Graph Convolutional Neural Network II (DGCNNII) model for graph classification task, which is up to 32 layers deep. And the effectiveness of our proposed method is demonstrated by quantifying the graph smoothness of each layer and ablation studies. Experiments on benchmark graph classification datasets show that DGCNNII outperforms a large number of shallow graph neural network baseline methods.

The role of internally-generated network dynamics in rapid temporal sequence coding, updating, and parallel recalling of alternate spatial and mental navigation contexts has remained unclear. Here, we revealed rapid emergence of temporally-compressed hippocampal theta sequences in adult male rats within 1-2 laps of a novel detour via re-purposing of pre-existing correlated neuronal sequence motifs expressed during pre-detour sleep. Detour experience-induced neuronal remapping and plasticity were consolidated and reconfigured hippocampal network during post-detour sleep, which predicted future representational drift expressed during the following post-detour reversal-track runs. Pre-detour or reversal-track representations flickered with detour representation across distinct phases of theta oscillation, revealing segregation of non-local internally-generated/recalled and local experience-related hippocampal representations by theta phase. These findings demonstrate that internal generative network dynamics across brain states support rapid theta-scale sequential coding, within-day representational updating, and flickering parallel representations — collectively forming non-local representations — during navigation. How the interplay between internal and external network drives shapes the representation of the world is not fully understood. The authors show that internal generative network dynamics support non-local theta sequence coding, representational updating and flickering during navigation.

Variable angle structural surfaces are more common in slopes. However, variable-angle joints have still been approximated as single-angle straight joint planes for analysis in previous studies, which failed to comprehensively capture the mechanical properties of such surfaces. In this paper, joint specimens with variable angles were fabricated using concrete. Direct shear tests were carried out under different normal stress conditions, and the strength, deformation, and failure characteristics of the specimens were analyzed. The results revealed a significant positive correlation between the peak shear strength and shear stiffness of the variable angle specimens with the normal stress, while the correlation between the residual shear strength and the normal stress was relatively insignificant. The normal displacement and tangential displacement of the specimen exhibited a significant negative correlation. When shear failure occurred, the main cracks propagated along the shear direction, accompanied by localized secondary cracks and surface spalling. Unlike the straight joints commonly studied in previous research, variable-angle joint specimens exhibit distinct failure modes in the high-angle zone and the low-angle zone. In the high-angle zone, vertical cracks perpendicular to the joint surface and shear cracks along the shear direction predominantly occurred. The low-angle zone was dominated by shear damage along the shear direction. By comparing the results of experimental tests and theoretical analysis, the zonal pattern of shear failure characteristics on variable angle surfaces was revealed. The research findings provide a theoretical basis for the safety and stability analysis of variable angle structural surfaces.

Children often learn the meanings of nouns before they grasp the meanings of verbs. This discrepancy could arise from differences in the complexity of visual characteristics for categories that language describes, the inherent structure of language, or how these two sources of information align. To explore this question, we analyze visual and linguistic representations derived from large-scale pre-trained artificial neural networks of common nouns and verbs, focusing on these three hypotheses about early verb learning. Our findings reveal that verb representations are more variable and less distinct within their domain compared to nouns. When only one example per category is available, the alignment between visual and linguistic representations is weaker for verbs than for nouns. However, with multiple examples (mirroring human language development), this alignment improves significantly for verbs, approaching that of nouns. This suggests that the difficulty in learning verbs is not primarily due to mapping visual events to verb meanings, but rather in forming accurate representations of each verb category. Regression analysis indicates that visual variability significantly impacts verb learning, followed by the alignment of visual and linguistic elements and linguistic variability. Our study provides a quantitative and integrative framework to account for the challenges children face in early word learning, opening new avenues for resolving the longstanding debate on why verbs are harder to learn than nouns.

Maize is one of the main food crops in the world, and cultivating high-yield and high-quality maize varieties is of great significance in addressing food security issues. Leaves are crucial photosynthetic organs in maize, and leaf senescence can result in the degradation of chlorophyll. This, in turn, impacts photosynthetic activity and the accumulation of photosynthetic products. Delaying leaf senescence and increasing carbon assimilation can enhance grain yield and biomass production. The stay green of maize is an important trait closely related to yield, feed quality and resistance. Therefore, this study employed multi-generation joint analysis of major genes and a polygene model to investigate the genetic inheritance of stay green-related traits. Four populations (P 1 , P 2 , F 1 and F 2 ) were obtained by crossing T01 (stay green) × Xin3 (non-stay green) and T01 (stay green) × Mo17 (non-stay green) under two environments. Six stay green-related traits, including visual stay green (VSG), number of green leaves (GLNM), SPAD value of ear leaf at anthesis (SPADS), SPAD value of ear leaf at maturity (SPADM), absolute green leaf area (GLAD), grain yield per plant (GYP), displayed continuous variations with kurtosis and skewness values of absolute value less than 1 and distribution close to normal. They were characterized by typical inheritance of quantitative traits, with these traits demonstrating the transgressive segregation. The correlation analysis among the traits revealed that five stay green traits have a positive impact on yield. VSG, GLNM and SPADM in the two populations were regulated by the two major genes of additive effects plus additive-dominance polygene model with a major gene heritability varying from 89.03 to 95.95% in the F 2 generation. GLAD in TMF 2 was controlled by two major genes of equal-additive dominance effects with high heritability (93.47%). However, in TXF 2 , GLAD was regulated by two major genes of additive-dominance interaction effects plus additive-dominance polygene model. These results provide important genetic information for breeding, which could guide the improvement of stay green-related traits. They also lay a foundation for quantitative trait loci mapping of the stay stay-green traits in maize.

Graph neural networks (GNNs) have significant advantages in dealing with non-Euclidean data and have been widely used in various fields. However, most of the existing GNN models face two main challenges: (1) Most GNN models built upon the message-passing framework exhibit a shallow structure, which hampers their ability to efficiently transmit information between distant nodes. To address this, we aim to propose a novel message-passing framework, enabling the construction of GNN models with deep architectures akin to convolutional neural networks (CNNs), potentially comprising dozens or even hundreds of layers. (2) Existing models often approach the learning of edge and node features as separate tasks. To overcome this limitation, we aspire to develop a deep graph convolutional neural network learning framework capable of simultaneously acquiring edge embeddings and node embeddings. By utilizing the learned multi-dimensional edge feature matrix, we construct multi-channel filters to more effectively capture accurate node features. To address these challenges, we propose the Co-embedding of Edges and Nodes with Deep Graph Convolutional Neural Networks (CEN-DGCNN). In our approach, we propose a novel message-passing framework that can fully integrate and utilize both node features and multi-dimensional edge features. Based on this framework, we develop a deep graph convolutional neural network model that prevents over-smoothing and obtains node non-local structural features and refined high-order node features by extracting long-distance dependencies between nodes and utilizing multi-dimensional edge features. Moreover, we propose a novel graph convolutional layer that can learn node embeddings and multi-dimensional edge embeddings simultaneously. The layer updates multi-dimensional edge embeddings across layers based on node features and an attention mechanism, which enables efficient utilization and fusion of both node and edge features. Additionally, we propose a multi-dimensional edge feature encoding method based on directed edges, and use the resulting multi-dimensional edge feature matrix to construct a multi-channel filter to filter the node information. Lastly, extensive experiments show that CEN-DGCNN outperforms a large number of graph neural network baseline methods, demonstrating the effectiveness of our proposed method.

Traditional villages, rich in historical and cultural value, hold a high level of preservation value. In the process of urbanization, traditional villages face the crisis of decline, making it difficult to perpetuate the carried cultural heritage. The Ganjiang River Basin hosts numerous traditional villages with rich research value, making the study of their preservation and development in this region a significant topic. This paper, from the perspective of landscape ecology, employs geographic detectors to analyze the driving factors behind the emergence of traditional villages in the Ganjiang River Basin, summarizing the spatial distribution characteristics of traditional villages. A classification method based on village landscape features is adopted to categorize traditional villages in the Ganjiang River Basin, providing a reference for planning the preservation and development of traditional villages. The research results show that plain areas are more suitable for the continuation of traditional villages; a single suitable environmental element cannot provide an environment conducive to the development of traditional villages, which is the result of the combined effect of multiple suitable elements; the study has divided traditional village landscapes into nine types, with clear distribution differences among different types of villages; for different regions and types of traditional villages, it is necessary to balance development and protection tendencies and plan differently according to environmental characteristics.

GSK2606414 is a new, effective, highly selective PERK inhibitor with adenosine-triphosphate-competitive characteristics. It can inhibit endoplasmic reticulum stress and has the possibility of treating periodontitis. However, owing to its strong hydrophobicity and side effects, highly efficient pharmaceutical formulations are urgently needed to improve the bioavailability and therapeutic efficacy of GSK2606414 in the treatment of periodontitis. Herein, a novel local GSK2606414 delivery system was developed by synthesizing GSK2606414 nanoparticles (NanoGSK) and further loading NanoGSK into a collagenase-responsive hydrogel. The drug release results showed that the drug-loaded hydrogels had outstanding enzymatic responsive drug release profiles under the local microenvironment of periodontitis. Furthermore, in vitro studies showed that the drug-loaded hydrogel exhibited good cellular uptake and did not affect the growth and proliferation of normal cells, while the drug-loaded hydrogel significantly improved the osteogenic differentiation of inflammatory cells. In the evaluations of periodontal tissue repair, the drug-loaded hydrogels showed a great effect on inflammation inhibition, as well as alveolar bone regeneration. Therefore, this work introduces a promising strategy for the clinical treatment of periodontitis.

Variable thickness composite laminates (VTCL) are susceptible to impact during use and may result in irreparable internal damage. In order to locate the internal impact damage of complex composite structures and monitor the impact signals of VTCL at the same time, a low velocity impact (LVI) monitoring system based on an optical fiber sensing network was constructed. Fiber Bragg grating (FBG) sensors are suitable for monitoring strain characteristics. By arranging FBG sensors on the laminate, we studied the spectrum analysis and localization of the impact signal collected by a FBG demodulator at constant temperature. The prior knowledge of variable thickness composite structures is difficult to obtain, and the multi-sensor dynamic monitoring is complex and difficult to realize. In order to locate the LVI of composite structures without prior knowledge, based on empirical mode decomposition (EMD), we proposed an impact localization method with zero-mean normalized cross-correlation (ZNCC) and thickness correction. The experimental results of LVI localization verification show that the ZNCC algorithm can effectively remove the temperature cross-sensitivity and impact energy influencing factors, and the thickness correction can reduce the interference of variable thickness characteristics on localization performance. The maximum localization error is 24.41 mm and the average error is 15.67 mm, which meets engineering application requirements. The method of variable-thickness normalization significantly improves impact localization performance for VTCL.

With the rapid development of Internet‐of‐Things (IoT) and mobile devices, the IoT applications become more computation‐intensive and latency‐sensitive, which bring severe challenges to the resource‐limited devices. Mobile Edge Computing has served as a key promising method to enhance the network's computing capability by enabling resource‐constrained devices to offload tasks to the edge servers. A major challenge, which has been overlooked by most existing works on task offloading, is the dependencies among tasks and subtasks. In this paper, the subtask offloading with logical dependency for IoT applications is focused on. Specifically, subtask dependent graphs are employed to explore the dependency of subtasks and consider the priority of task scheduling. Further, an offloading scheme is put forward for minimizing both task latency and energy consumption of the device with dependency guarantees for all IoT tasks in multi‐server edge networks. Finaly, the simulation results demonstrate that the overall reduction rate is around 14% and relatively stable can effectively reduce task latency in multi‐server edge networks.