Complex Network Structure Research Articles

JotlasNet: Joint tensor low-rank and attention-based sparse unrolling network for accelerating dynamic MRI.

Joint low-rank and sparse unrolling networks have shown superior performance in dynamic MRI reconstruction. However, existing works mainly utilized matrix low-rank priors, neglecting the tensor characteristics of dynamic MRI images, and only a global threshold is applied for the sparse constraint to the multi-channel data, limiting the flexibility of the network. Additionally, most of them have inherently complex network structure, with intricate interactions among variables. In this paper, we propose a novel deep unrolling network, JotlasNet, for dynamic MRI reconstruction by jointly utilizing tensor low-rank and attention-based sparse priors. Specifically, we utilize tensor low-rank prior to exploit the structural correlations in high-dimensional data. Convolutional neural networks are used to adaptively learn the low-rank and sparse transform domains. A novel attention-based soft thresholding operator is proposed to assign a unique learnable threshold to each channel of the data in the CNN-learned sparse domain. The network is unrolled from the elaborately designed composite splitting algorithm and thus features a simple yet efficient parallel structure. Extensive experiments on two datasets (OCMR, CMRxRecon) demonstrate the superior performance of JotlasNet in dynamic MRI reconstruction.

An Evolutionary Game Model of a Regional Logistics Service Supply Chain Complex Network in a Blockchain Environment

Blockchain technology offers a novel solution to address the issue of information silos present in the complex network system of regional logistics service supply chains. Furthermore, enhancing the willingness of participants to engage in this system has emerged as a critical issue requiring urgent resolution. This study assumes bounded rationality and constructs an evolutionary game model to analyze the adoption behaviors of logistics service providers (LSPs) and logistics service integrators (LSIs) in a blockchain environment. The stability theorem of differential equations is employed to elucidate the inherent decision-making mechanisms present in the game. Acknowledging that traditional evolutionary mechanisms often neglect individual heterogeneity and the dynamic changes in the topological structures of complex networks, this study further involves the development of an evolutionary game model tailored for the complex network of a regional logistics service supply chain. After validating the effectiveness of the model, this research investigates how market environmental factors influence the adoption behaviors of LSPs and LSIs. Finally, through case simulations, the effects of key influencing factors on participatory behaviors within the regional logistics service supply chain are analyzed. The findings indicate that, in a blockchain environment, both prior to and following the adoption of complex network technology in the regional logistics service supply chain, the equilibrium stability of the network system is jointly determined by the total revenues of LSPs and LSIs. External factors in the blockchain environment significantly influence technology adoption strategies. Notably, when both the sensitivity coefficients of market demand relative to price and logistics service costs concerning quality enhancement are high, the willingness of supply chain participants to adopt the technology declines rapidly. As the quality elasticity increases, the technology adoption rates of providers and integrators initially stabilize and subsequently increase exponentially. Changes in the revenues of logistics service providers have no significant impact on technology adoption rates within the regional logistics service supply chain.

Identifying vital edges based on the cycle structure in complex networks

Identifying vital edges based on the cycle structure in complex networks

Detectability constraints on meso-scale structure in complex networks.

Community, core-periphery, disassortative and other node partitions allow us to understand the organisation and function of large networks. In this work we study common meso-scale structures using the idea of block modularity. We find that the configuration model imposes strong restrictions on core-periphery and related structures in directed and undirected networks. We derive inequalities expressing when such structures can be detected under the configuration model which are closely related to the resolution limit. Nestedness is closely related to core-periphery and is similarly restricted to only be detectable under certain conditions. We then derive a general equivalence between optimising block modularity and maximum likelihood estimation of the parameters of the degree corrected Stochastic Block Model. This allows us to contrast the two approaches, how they formalise the structure detection problem and understand these constraints in inferential versus descriptive approaches to meso-scale structure detection.

New insights into autochthonous fungal bioaugmentation mechanisms for recalcitrant petroleum hydrocarbon components using stable isotope probing.

Autochthonous fungal bioaugmentation (AFB) is a promising strategy for the microbial remediation of petroleum hydrocarbon (PH)-contaminated soils. However, the mechanisms underlying AFB, particularly for degrading recalcitrant PH components, are not fully understood. This study employed stable isotope probing (SIP) and high-throughput sequencing to investigate the AFB mechanisms of two hydrocarbon-degrading fungi, Fusarium solani LJD-11 and Aspergillus fumigatus LJD-29, focusing on three challenging PH components: n-Hexadecane (n-Hex), Benzo[a]pyrene (BaP), and Dibenzothiophene (DBT). Our findings indicate that both fungal strains significantly enhanced pollutant removal rates, with combined application yielding optimal results. AFB treatment reduced the microbial diversity index and altered the soil microbial community, especially affecting fungal populations. A significant correlation between the microbial diversity index and degradation efficiency suggests that greater diversity enhances pollutant removal. SIP analysis showed that LJD-11 and LJD-29 could directly assimilate n-Hex and DBT, but not BaP. Correlation analyses between functional microorganisms and other biological indicators suggest that the removal of pollutants is also attributable to indigenous functional bacteria. Additionally, non-inoculated functional fungi present in the soil play a crucial role in BaP degradation. These findings reveal distinct degradation pathways for the three pollutants. The addition of carrier substrate reduced the complexity of the network, while AFB treatment restored it. In addition, the combined fungal treatment resulted in higher network parameters, leading to a more complex and stable network structure. These results provide insights into the mechanisms of AFB for degrading recalcitrant PH components, underscoring its potential for in situ bioremediation of petroleum-contaminated soils.

Niche partitioning in a periphyton metacommunity peaks at intermediate species richness in midsized rivers.

The trait-based partitioning of species plays a critical role in biodiversity-ecosystem function relationships. This niche partitioning drives and depends on community structure, yet this link remains elusive in the context of a metacommunity, where local community assembly is dictated by regional dispersal alongside local environmental conditions. Hence, elucidating the coupling of niche partitioning and community structure needs spatially explicit studies. Such studies are particularly necessary in river networks, where local habitats are highly connected by unidirectional water flow in a spatially complex network structure and frequent disturbance makes community structure strongly dependent on recolonization. Here, we show that taxonomic turnover among periphyton communities colonizing deployed bricks (microhabitats) at multiple sampling sites (local habitats) in a river network came along with a turnover in traits. This niche partitioning showed a hump-shaped relationship with richness of periphyton communities, which increased along river size. Our observations suggest downstream dispersal along the river network to increase the regional metacommunity pool, which then ensures local colonization by taxa possessing diverse traits allowing them to efficiently partition into environmentally different microhabitats. However, at the most downstream sites, the excessive dispersal of widespread generalists drove mass effects which inflated richness with taxa that co-occupied several microhabitats and swamped niche partitioning. Further, efficient niche partitioning depended on communities rich in rare taxa, an indication for the importance of specialists. Alarmingly, richness and rare taxa declined with high phosphorus concentrations and conductivity, respectively, two environmental variables which potentially reflected anthropogenic activity.

The Role of Graph-Based Data Science Tools in Uncovering Complex Network Relationships

Graph-based data science tools have emerged as essential methodologies for analyzing and interpreting complex network structures prevalent in a wide range of domains. These tools enable researchers and practitioners to discover hidden patterns, understand community dynamics, identify influential nodes, and reveal intricate relational structures that would otherwise remain elusive. This paper examines the role of graph-based data science approaches in uncovering complex network relationships by exploring key computational frameworks, analytical methods, and visualization techniques. The work delves into the theoretical foundations of graph representations, evaluates the existing literature on graph analytics frameworks, and illustrates how these tools are implemented across various contexts, including social networks, biological interactions, financial markets, and communication networks. Through an empirical methodology, this paper evaluates tool performance and introduces a case study where a real-world social media interaction network is analyzed to identify latent communities and measure node centrality. The results demonstrate that graph-based data science tools allow for more efficient and meaningful exploration of network properties, contributing not only to a more profound understanding of underlying relational structures but also to improved decision-making processes and strategic planning. The paper concludes by underscoring the importance of continued research and development of graph analytics frameworks to better support the evolving complexity and scale of modern data.

Identifying influential individuals in social networks: An example of a location-based online social network

The landscape of information access has evolved significantly over time, with the advent of search engines, social media platforms, and the widespread use of the internet. These developments have fostered a global communication network, resulting in intricate connections between individuals. Online social networks have emerged as key facilitators of social interaction, expediting the exchange of information and playing a pivotal role in content dissemination. Within these networks, certain individuals, termed as Key Players, wield considerable influence, profoundly impacting information diffusion. Thus, the identification of the most influential individuals within complex network structures stands as a crucial challenge. In this study, we employ modularity and eigenvector centrality metrics to designate nodes for initial activation, aiming at influence maximization in social networks. Visualization and analysis of the dataset are conducted using Gephi software, providing insights into the dynamics of the social network structure and facilitating the identification of key players.

Rice Disease Classification Using a Stacked Ensemble of Deep Convolutional Neural Networks

Rice is a staple food for almost half of the world’s population, and the stability and sustainability of rice production plays a decisive role in food security. Diseases are a major cause of loss in rice crops. The timely discovery and control of diseases are important in reducing the use of pesticides, protecting the agricultural eco-environment, and improving the yield and quality of rice crops. Deep convolutional neural networks (DCNNs) have achieved great success in disease image classification. However, most models have complex network structures that frequently cause problems, such as redundant network parameters, low training efficiency, and high computational costs. To address this issue and improve the accuracy of rice disease classification, a lightweight deep convolutional neural network (DCNN) ensemble method for rice disease classification is proposed. First, a new lightweight DCNN model (called CG-EfficientNet), which is based on an attention mechanism and EfficientNet, was designed as the base learner. Second, CG-EfficientNet models with different optimization algorithms and network parameters were trained on rice disease datasets to generate seven different CG-EfficientNets, and a resampling strategy was used to enhance the diversity of the individual models. Then, the sequential least squares programming algorithm was used to calculate the weight of each base model. Finally, logistic regression was used as the meta-classifier for stacking. To verify the effectiveness, classification experiments were performed on five classes of rice tissue images: rice bacterial blight, rice kernel smut, rice false smut, rice brown spot, and healthy leaves. The accuracy of the proposed method was 96.10%, which is higher than the results of the classic CNN models VGG16, InceptionV3, ResNet101, and DenseNet201 and four integration methods. The experimental results show that the proposed method is not only capable of accurately identifying rice diseases but is also computationally efficient.

MGFusion: a multimodal large language model-guided information perception for infrared and visible image fusion

Existing image fusion methods primarily focus on complex network structure designs while neglecting the limitations of simple fusion strategies in complex scenarios. To address this issue, this study proposes a new method for infrared and visible image fusion based on a multimodal large language model. The method proposed in this paper fully considers the high demand for semantic information in enhancing image quality as well as the fusion strategies in complex scenes. We supplement the features in the fusion network with information from the multimodal large language model and construct a new fusion strategy. To achieve this goal, we design CLIP-driven Information Injection (CII) approach and CLIP-guided Feature Fusion (CFF) strategy. CII utilizes CLIP to extract robust image features rich in semantic information, which serve to supplement the information of infrared and visible features, thereby enhancing their representation capabilities for the scene. CFF further utilizes the robust image features extracted by CLIP to select and fuse the infrared and visible features after the injection of semantic information, addressing the challenges of image fusion in complex scenes. Compared to existing methods, the main advantage of the proposed method lies in leveraging the powerful semantic understanding capabilities of the multimodal large language model to supplement information for infrared and visible features, thus avoiding the need for complex network structure designs. Experimental results on multiple public datasets validate the effectiveness and superiority of the proposed method.

Network Embedding Methods: Study and Comparison

Graph is a powerful language that can model many systems in different fields such as information sciences, social sciences, Biology, mathematics, physics, etc. Graphs can capture very well the relationships between nodes and their structure. Representing data through graphs has some limitations and is challenging to use them like input in machine learning and deep learning models. This challenge can be overcome by using network embedding. Embedding represents a network into low-dimensional vector space. Several methods have been proposed to embed networks. Methods like DeepWalk, Node2Vec and GraphSAGE have gained traction. DeepWalk employs random walks to capture local neighborhood information, generating embeddings using skip-gram models. Node2Vec extends this by incorporating biased random walks, balancing exploration and exploitation of the network. GraphSAGE adopts a neighborhood aggregation strategy, aggregating features from a node's local graph structure through different convolutional layers. These methods facilitate downstream tasks like node classification, link prediction and clustering. Challenges include scalability to large networks, handling heterogeneous information and preserving structural characteristics. Recent advancements integrate techniques like attention mechanisms and reinforcement learning for improved embeddings. Network embedding methods continue to evolve, catering to diverse applications in social networks, biology and recommendation systems, offering insights into complex network structures. In this paper we provide networks embedding concepts and some methods. A comparison between embedding methods is presented. Embedding network methods on node classification tasks on different benchmark datasets have been conducted in order to evaluate them. Through rigorous evaluation, we assess their effectiveness in capturing latent features and improving classification accuracy across various domains.

Synthetic graphs for link prediction benchmarking

Abstract Predicting missing links in complex networks requires algorithms that are able to explore statistical regularities in the existing data. Here we investigate the interplay between algorithm efficiency and network structures through the introduction of suitably-designed synthetic graphs. We propose a family of random graphs that incorporates both micro-scale motifs and meso-scale communities, two ubiquitous
structures in complex networks. A key contribution is the derivation of theoretical upper bounds for link prediction performance in our synthetic graphs, allowing us to estimate the predictability of the task and obtain an improved assessment of the performance of any method. Our results on the performance of classical methods (e.g., Stochastic Block Models, Node2Vec, GraphSage) show that the performance of all methods correlate with the theoretical predictability, that no single method is universally superior, and that each of the methods exploit different characteristics known to exist in large classes of networks. Our findings underline the need for careful consideration of graph structure when selecting a link prediction method and emphasize the value of comparing performance against synthetic benchmarks. We provide open-source code for generating these synthetic graphs, enabling further research on link
prediction methods.

Prediction of human disease complication incorporating machine learning and local network structures

Abstract Complications have long haunted physicians and patients in clinical medicine. However, the evaluation of complications caused by specific diseases is typically relied on the experience of clinicians or clinical cases. Especially, vast complication diseases involve multiple human body systems, increasing the difficulty of the clinical confirmation. Based on a large scale human disease complication network extracted from the clinical medicine knowledge database, we propose a nonlinear model combined local topological structures and machine learning to explore latent disease-complication relations. As an example, we apply the model to predict unidentified complications of COVID-19 and to detect potential extrapulmonary complications which are significant in the post-pandemic period. Our approach provides an efficient method to identify the candidate complications from the structure of complex network.

Regulation of thermal transport by cycle structures in complex networks

The regulation of thermal transport in complex networks is a challenging topic. This paper focuse on thermal transport in the Erdős-Rényi network by adjusting two coefficients: the number of cycle structures and the cycle ratio. The results indicate that the increase of cycles in Erdős-Rényi networks leads to enhanced phonon localization and reduced efficiency in thermal transport. Further more, as the average cycle ratio increases, we observe an improvement of the network’s thermal transport efficiency, accompanied by thermal rectification phenomena. Finally, through the analysis of participation ratio, eigenvalue spectra, and node vibration time series, we clarified that the variations in phonon localization and the widths of the phonon spectrum are key factors influencing the thermal transport in complex networks. This work enables control of thermal transport by adjusting the structural characteristics of the network and provides further exploration and development of phonon localization in disordered systems.

An ultra-lightweight network combining Mamba and frequency-domain feature extraction for pavement tiny-crack segmentation

Pixel-level detection and segmentation of pavement cracks is a crucial but challenging task. Current deep learning methods mainly target larger cracks, often overlooking tiny cracks. Additionally, improvements in these methods typically involve more complex network structures, leading to reduced efficiency and higher computational demands. To tackle these challenges, we propose an Ultra-lightweight Network (ULNet) designed for tiny crack segmentation. ULNet boasts a low parameter count and minimal FLOPs, comprising three main components: the Mamba feature extraction branch, the frequency domain feature extraction branch, and the feature decoder. The Mamba feature extraction branch utilizes the proposed cross-visual Mamba feature extraction module, which enhances the accuracy of tiny crack detection by extracting global features through layer-by-layer processing and cross-fusion of feature information, all without relying on an attention mechanism. The frequency domain feature extraction branch transforms input features into the frequency domain and applies a Gaussian high-pass filter to smooth the image and extract high-frequency crack details while suppressing noise. Finally, the feature decoder incorporates a multi-scale feature selection and fusion module that integrates both channel and spatial attention, enabling effective extraction of tiny crack features from adjacent scales. Inspired by the mechanisms of biological visual information processing, we integrate the high-frequency features from the frequency domain feature extraction branch with the features from the Mamba feature extraction branch. This integration follows the interactive patterns of biological vision, allowing for layer-wise fusion and interaction of features, thereby supplementing the information from the Mamba branch and reducing information loss. We tested and evaluated ULNet on several public datasets (BJN260, Rain365, Sun520, and DeepCrack). Experimental results indicate that our method achieves state-of-the-art performance among all lightweight approaches.

Hybrid game model for electricity trading and pricing among multiple microgrids and consumers based on demand-side complex networks

Local energy markets (LEMs) are pivotal for enhancing renewable energy consumption and facilitating green electricity by linking multiple microgrids (MMGs) and consumers. Existing studies often overlook the evolution of consumer purchasing strategies influenced by social spatial structures and social learning dynamics. This paper addresses this gap by introducing a hybrid game model for LEMs electricity trading and dynamic pricing, integrating demand-side complex network structures. This model encompasses dynamic interactions between MMGs and electricity and carbon markets, employing non-cooperative games among MMGs, complex network evolutionary game among consumers, and Stackelberg games between MMGs and consumers. It promotes efficient transactions and adaptive pricing through distributed algorithms optimizing equilibrium solutions. Key findings include: (1) Compared to Power-to-Grid and Peer-to-Peer models, our model increased MMGs' revenues by 44.10 % and 10.60 %, reduced consumers’ costs by 8.76 % and 1.15 %, and cut carbon emissions by 93.32 % and 77.30 %, significantly boosting economic and environmental benefits. (2) Complex network analysis underscores the importance of social learning in adjusting power consumption strategies and pricing mechanisms. (3) Multidimensional consumer decision-making enhances renewable energy valuation, fosters price competition among MMGs, and accelerates low-carbon transitions in LEMs, providing crucial theoretical support for LEM structure design and policy to enhance renewable energy use.

Fungal degradation of complex organic carbon supports denitrification in saturated woodchip bioreactors

Woodchip bioreactor (WBR) is a promising technology for the removal of nitrate from agricultural drainage, although the performance of WBRs is dependent on the decomposition of lignocellulosic biomass and the carbon availability for microbial denitrification. Fungal species are more efficient than bacterial counterparts in driving wood decomposition; however, little is known about the fungal community structure and functions in saturated WBRs. In this study, we investigated the dynamics of the mycobiome in field-scale, constantly saturated WBRs located in Willmar, Minnesota, USA. Fungal community analysis suggested that wood-rotting fungi were abundant in WBRs, especially near their inlet locations where microbial denitrification was most active. Complex network structures of fungal hyphae associated with a decayed cavity on the woodchip surface was further evidenced by confocal and scanning electron microscopy. These results suggest that fungi play a major role in wood degradation in WBRs, thereby promoting denitrification activity.

RJ-TinyViT: an efficient vision transformer for red jujube defect classification

Compared to the surface defect detection of industrial products produced according to specified processes, the detection of surface defects in naturally grown red jujubes poses unique and significant challenges for researchers. The high diversity of surface defects, subtle distinctions from the background, low contrast, varying scales, and the presence of high levels of noise in images are among the factors that greatly amplify the complexity of defect detection tasks. Existing methods show some deficiencies in addressing these issues, mainly due to insufficient feature extraction capabilities and overly complex network structures, leading to limitations in model efficiency and practical application performance. To tackle the challenges associated with red jujube surface defect detection, this study proposes an optimized Tiny Vision Transformer (TinyViT) network structure, named RJ-TinyViT. This method refines the TinyViT-5 m network structure to reduce network burden and introduces an improved Multi-Kernel Block (MK Block) and an improved Mobile Inverted Bottleneck Convolution Block (MBConv Block) to enhance feature extraction capabilities. Additionally, we have integrated the Coordinate Attention (CA) module to enhance the model’s capacity for recognizing and focusing on features of surface defects on red jujubes. Experimental results show that RJ-TinyViT achieved a classification accuracy of 93.38%, marking an improvement of 1.84% over the original TinyViT network. At the same time, its Floating-point Operations (FLOPs) and Parameters (Params) were reduced to 58.97% and 39.84% of the original TinyViT network, respectively. These results not only demonstrate that RJ-TinyViT achieves model lightweighting while maintaining high accuracy but also highlight its value in practical industrial applications.

Residual trio feature network for efficient super-resolution

Deep learning-based approaches have demonstrated impressive performance in single-image super-resolution (SISR). Efficient super-resolution compromises the reconstructed image’s quality to have fewer parameters and Flops. Ensured efficiency in image reconstruction and improved reconstruction quality of the model are significant challenges. This paper proposes a trio branch module (TBM) based on structural reparameterization. TBM achieves equivalence transformation through structural reparameterization operations, which use a complex network structure in the training phase and convert it to a more lightweight structure in the inference, achieving efficient inference while maintaining accuracy. Based on the TBM, we further design a lightweight version of the enhanced spatial attention mini (ESA-mini) and the residual trio feature block (RTFB). Moreover, the multiple RTFBs are combined to construct the residual trio network (RTFN). Finally, we introduce a localized contrast loss for better applicability to the super-resolution task, which enhances the reconstruction quality of the super-resolution model. Experiments show that the RTFN framework proposed in this paper outperforms other state-of-the-art efficient super-resolution methods in terms of inference speed and reconstruction quality.

Effects of burning on vegetation, soil physicochemistry and prokaryotic microbial communities in surface and subsurface peat

Prescribed burning is a common management strategy in peatlands that has the potential to affect soil physicochemistry, alter biogeochemical cycles and trigger changes in vegetation structure. How burning affects prokaryotic community composition across different soil profiles is not well understood. This study explored the effects of prescribed burning on the diversity of prokaryotic communities in peat soils. Soil samples were collected from Moor House Nature Reserve, UK, a long-term monitoring site initiated in 1954 subject to three burning treatments: Burning at short rotations every 10 years, burning at long rotations every 20 years and a non-burn control. Observed species richness for archaea was highest in the topsoil of the non-burn control plots and highest for bacteria in the topsoil of the non-burn control and plots under a long rotation regime. Community composition was significantly different between different burn treatments and soil depth. Archaeal community structure was shaped by NH4+ and pH in the topsoil; by Pb, moisture and Al in the 20–40 cm profile; and by total N, total C, Al, Ca, Fe and pH in the 40–60 cm profile. Bacterial community structure was shaped by NH4+, heather cover, pH and Mg in the topsoil; by Fe, K and Pb in the 20–40 cm profile; and by Al, Ca and Fe in the 40–60 cm profile. A co-occurrence network analysis revealed that the topsoil of the non-burn control plots had a more complex network structure with more positive links than those under a rotational burn, but a higher average connectivity with a higher number of negative links was observed in the long rotation 20–40 cm profile. The results provide a new insight into the response processes of soil prokaryotic communities to burning in peatland soils, providing valuable knowledge that can support the evaluation and management of ecosystem services in peatlands.