Complex Network Structure Research Articles

Radar Human Activity Recognition with an Attention-Based Deep Learning Network

Radar-based human activity recognition (HAR) provides a non-contact method for many scenarios, such as human-computer interaction, smart security, and advanced surveillance with privacy protection. Feeding radar-preprocessed micro-Doppler signals into a deep learning (DL) network is a promising approach for HAR. Conventional DL algorithms can achieve high performance in terms of accuracy, but the complex network structure causes difficulty for their real-time embedded application. In this study, an efficient network with an attention mechanism is proposed. This network decouples the Doppler and temporal features of radar preprocessed signals according to the feature representation of human activity in the time-frequency domain. The Doppler feature representation is obtained in sequence using the one-dimensional convolutional neural network (1D CNN) following the sliding window. Then, HAR is realized by inputting the Doppler features into the attention-mechanism-based long short-term memory (LSTM) as a time sequence. Moreover, the activity features are effectively enhanced using the averaged cancellation method, which improves the clutter suppression effect under the micro-motion conditions. Compared with the traditional moving target indicator (MTI), the recognition accuracy is improved by about 3.7%. Experiments based on two human activity datasets confirm the superiority of our method compared to traditional methods in terms of expressiveness and computational efficiency. Specifically, our method achieves an accuracy close to 96.9% on both datasets and has a more lightweight network structure compared to algorithms with similar recognition accuracy. The method proposed in this article has great potential for real-time embedded applications of HAR.

Provincial Inclusive Green Growth Efficiency in China: Spatial Correlation Network Investigation and Its Influence Factors

Inclusive green growth efficiency (IGGE) analysis is an effective tool for improving coordinated economic, social, and environmental development. This study incorporated the game cross-efficiency DEA to measure the IGGE of 30 provinces in China. Then, the modified spatial gravity model and social network analysis model were applied to construct and analyze the spatial correlation network structure of the IGGE. The quadratic assignment procedure was used to mine the influencing factors that affect the formation and evolution of the spatial correlation network of the IGGE. The results are as follows. (1) During the study period, there were significant differences in the IGGE among the 31 provinces, among which the eastern provinces were higher than the central and western provinces. (2) The spatial correlation of the IGGE presented a complex and multi-threaded network structure, indicating that the IGGE has a noticeable cross-regional spillover effect. Beijing, Tianjin, Zhejiang, Shanghai, Jiangsu, and Guangdong played the role of the “net spillover” block. Qinghai, Guizhou, Guangxi, and the surrounding provinces played the role of the “primary beneficial”. The Yangtze delta and Pearl River Delta economic zone (primarily including Shanghai and Guangdong) acted as a “bridge” to the Yunnan–Guizhou region and the surrounding provinces. (3) The spatial adjacency, degree of openness, economic development, and environmental governance were the prominent factors influencing the formation and evolution of the IGGE spatial correlation network. This work provides an example of constructing an IGGE correlation network while considering various factors, such as the economy, population, and distance. It also could help policymakers clarify the IGGE spatial correlation pattern and the provinces’ roles and potential for IGGE synergic improvement.

Lightweight Tennis Ball Detection Algorithm Based on Robomaster EP

To address the problems of poor recognition effect, low detection accuracy, many model parameters and computation, complex network structure, and unfavorable portability to embedded devices in traditional tennis ball detection algorithms, this study proposes a lightweight tennis ball detection algorithm, YOLOv5s-Z, based on the YOLOv5s algorithm and Robomater EP. The main work is as follows: firstly, the lightweight network G-Backbone and G-Neck network layers are constructed to reduce the number of parameters and computation of the network structure. Secondly, convolutional coordinate attention is incorporated into the G-Backbone to embed location information into channel attention, which enables the network to obtain location information of a larger area through multiple convolutions and enhances the expression ability of mobile network learning features. In addition, the Concat module in the original feature fusion is modified into a weighted bi-directional feature pyramid W-BiFPN with settable learning weights to improve the feature fusion capability and achieve efficient weighted feature fusion and bi-directional cross-scale connectivity. Finally, the Loss function EIOU Loss is introduced to split the influence factor of the aspect ratio and calculate the length and width of the target frame and anchor frame, respectively, combined with Focal-EIOU Loss to solve the problem of imbalance between complex and easy samples. Meta-ACON’s activation function is introduced to achieve an adaptive selection of whether to activate neurons and improve the detection accuracy. The experimental results show that compared with the YOLOv5s algorithm, the YOLOv5s-Z algorithm reduces the number of parameters and computation by 42% and 44%, respectively, reduces the model size by 39%, and improves the mean accuracy by 2%, verifying the effectiveness of the improved algorithm and the lightweight of the model, adapting to Robomaster EP, and meeting the deployment requirements of embedded devices for the detection and identification of tennis balls.

Global transfer of salinization on irrigated land: Complex network and endogenous structure

Increasing globalization intensifies land redistribution via global supply chains. Interregional trade not only transfers embodied land but also displaces the negative environmental impact of land degradation from one region to another. This study sheds light on land degradation transfer by focusing on salinization directly whereas previous studies have extensively assessed the land resource embodied in trade. To analyze the relationships among economies under interwoven embodied flows, this study integrates complex network analysis and input-output method to observe the endogenous structure of the transfer system. By focusing on irrigated land with higher crop yields than dryland farming, we make policy recommendations on food safety and proper irrigation. The results of the quantitative analysis show that the total amount of saline and sodic irrigated land embodied in global final demand are 260978.23 and 424291.05 square kilometers respectively. Salt-affected area of irrigated land is imported by not only developed countries but also large developing countries such as Mainland China and India. Exports of embodied salt-affected land in Pakistan, Afghanistan, and Turkmenistan are pressing issues, accounting for nearly 60% of total exports from net exporters worldwide. It is also demonstrated that embodied transfer network has a basic community structure of three groups due to regional preference in agricultural products trade.

Synthesis, characterization, and application of starch-based adhesives modified with itaconic acid and N-hydroxyethyl acrylamide

A renewable, biodegradable, formaldehyde-free starch-based adhesive was prepared by grafting itaconic acid (IA) and N-hydroxyethyl acrylamide (HEAA) onto corn starch. The adhesive exhibits excellent performance for particleboard and low viscosity (251 mPas, 62 s measured by flow cup viscometer) using 8% addition of IA (w/w, dry starch basis), and 10% addition of HEAA (w/w, dry starch basis). The thickness swelling of the particleboard was decreased to 10%. When unmodified starch-based adhesive was used, the particleboard soaked in water disintegrated completely. The data obtained from the Fourier transform infrared (FTIR) spectroscopy, scanning electron microscopy (SEM), confocal Raman microscopy (CRM), and 13C cross-polarization magic-angle spinning (CPMAS) showed that IA wrapped and covered the surface of starch granules. HEAA is unevenly distributed on the surface of starch as a cross-linked backbone. After the hot-pressing process, IA and HEAA are self-cross-linked and inter-cross-linked, respectively, to form a complex network structure, which can inhibit the disintegration caused by excessive expansion during the soaking process of the board, thus improving the water resistance of the starch adhesive.

MA-GCN: A Memory Augmented Graph Convolutional Network for traffic prediction

Traffic forecasting is a particularly challenging and important application direction in the field of spatial–temporal prediction. However, it is difficult for existing models to accurately capture the long time dependence of traffic data and the complex spatial dependence of road network. To solve these two issues, in this work, we propose a new deep learning framework — Memory Augmented Graph Convolutional Network (MA-GCN), which combines graph convolutional network (GCN) with differential neural computer (DNC). In the model, GCN is used to learn the complex road network structure to capture the spatial dependence, while DNC is applied to learn the long-term dynamic changes of traffic data to capture the long time dependence. Based on this, the traffic prediction is implemented, and the experimental evaluation is carried out on two public datasets, PeMSD4 and PeMSD8. The results show that the MA-GCN model is superior to the comparative models on several evaluation metrics.

Optimal Allocation and Sizing of Fault Current Limiters Considering Transmission Switching With an Explicit Short Circuit Current Formulation

As the power system expands with a more complex network structure, the excessive short-circuit current (SCC) has become a major obstacle to the rapid development of power system. The paper proposes a SCC-constrained planning model to determine the optimal locations and sizes of fault current limiters (FCLs) while considering transmission switching. N-1 security criterion is included to ensure that switching off lines will not jeopardize system security. One challenge is that the relationship between SCC and binary variables associated with switching states of transmission lines and activation statuses of FCLs is nonlinear and hard to obtain the exact expression. To this end, the paper begins with the equations of admittance matrix and nodal self-impedance, and derives the explicit formulation of SCC considering transmission switching and FCL activation. Based on the explicit expression of SCC, the proposed optimal allocation and sizing model of FCLs is formulated as a mixed integer linear programming (MILP) problem which could be solved by off-the-shelf commercial solver Gurobi in MATLAB. Numerical case studies on IEEE 30-bus, 118-bus and 1354-bus systems validate the effectiveness and advantages of the proposed model in reducing total costs and enhancing system security. The modeling accuracy of the explicit SCC formulation is also illustrated.

Anisotropic Strength, Deformability, and Failure Behavior of Artificial Columnar Jointed Rock Masses under Triaxial Compression

Columnar jointed rock masses (CJRMs) have a complex network structure and, therefore, present a great challenge to geotechnical engineering design. A series of triaxial compression tests is performed on artificial CJRM specimens subjected to a confining pressure of 5 MPa, considering various pore pressures (pw=0 MPa, 2 MPa, and 4 MPa). In this work, the influence of the joint dip angles and pore pressure on the anisotropic strength, deformation, and failure modes of the CJRM is investigated. The anisotropic characteristic stresses, strength parameters (c and φ) and elastic modulus of the CJRM specimens show a U-shaped pattern with increasing joint dip angles (β), while Poisson’s ratio changes inversely. The elastic modulus has a stronger anisotropy than Poisson’s ratio. Moreover, with increasing pore pressure, the characteristic stresses decrease, while the anisotropy of stresses and deformations increases. Finally, three failure modes of CJRM specimens with different joint dip angles are observed under triaxial compression conditions.

Synergistic effects of sugar beet pectin and laccase on gel properties of egg white

It is of great significance to improve the gel properties of egg white for their related products. This goal was accomplished by investigating the gel properties of egg white via interacting with pectin and laccase, and revealing their mechanisms. Results showed that 0.06% sugar beet pectin (SBP) and 0.5 U/g laccase could endow egg white gels with better texture properties (with the highest hardness of 1 508.00 g, chewiness of 1 144.83, and gumminess of 1 210.57), and superior water holding capacity (WHC) (the highest WHC value of 72.08%). This was attributed to the fact that 0.06% of SBP filled the pores of egg white gel, and enhanced affinity and binding of the gels to water by the good hydrophilicity of SBP, and the decrease in egg white gel pore size. Laccase (0.5 U/g) could promote the cross-linking between the egg white gel and SBP to form a more complex gel network structure through the interactions between the tyrosine residues and ferulic acid in SBP, which brought more ionic bonds (22.47%), hydrogen bonds (12.4%), disulfide bonds (16.97%) and more compact structure to egg white gels. This work can provide an effective method to improve the quality of egg white gel and related products.

Device identification using optimized digital footprints

<span lang="EN-US">The rapidly increasing number of internet of things (IoT) and non-IoT devices has imposed new security challenges to network administrators. Accurate device identification in the increasingly complex network structures is necessary. In this paper, a device fingerprinting (DFP) method has been proposed for device identification, based on digital footprints, which devices use for communication over a network. A subset of nine features have been selected from the network and transport layers of a single transmission control protocol/internet protocol packet based on attribute evaluators in Weka, to generate device-specific signatures. The method has been evaluated on two online datasets, and an experimental dataset, using different supervised machine learning (ML) algorithms. Results have shown that the method is able to distinguish device type with up to 100% precision using the random forest (RF) classifier, and classify individual devices with up to 95.7% precision. These results demonstrate the applicability of the proposed DFP method for device identification, in order to provide a more secure and robust network.</span>

Factors driving the halophyte rhizosphere bacterial communities in coastal salt marshes.

Root-associated microorganisms promote plant growth and provide protection from stresses. Halophytes are the fundamental components maintaining ecosystem functions of coastal salt marshes; however, it is not clear how their microbiome are structured across large spatial scales. Here, we investigated the rhizosphere bacterial communities of typical coastal halophyte species (Phragmites australis and Suaeda salsa) in temperate and subtropical salt marshes across 1,100 km in eastern China. The sampling sites were located from 30.33 to 40.90°N and 119.24 to 121.79°E across east China. A total of 36 plots were investigated in the Liaohe River Estuary, the Yellow River Estuary, Yancheng, and Hangzhou Bay in August 2020. We collected shoot, root, and rhizosphere soil samples. the number of pakchoi leaves, total fresh and dry weight of the seedlings was counted. The soil properties, plant functional traits, the genome sequencing, and metabolomics assay were detected. The results showed that soil nutrients (total organic carbon, dissolved organic carbon, total nitrogen, soluble sugars, and organic acids) are high in the temperate marsh, while root exudates (measured by metabolite expressions) are significantly higher in the subtropical marsh. We observed higher bacterial alpha diversity, more complex network structure, and more negative connections in the temperate salt marsh, which suggested intense competition among bacterial groups. Variation partitioning analysis showed that climatic, edaphic, and root exudates had the greatest effects on the bacteria in the salt marsh, especially for abundant and moderate subcommunities. Random forest modeling further confirmed this but showed that plant species had a limited effect. Taken together, the results of this study revealed soil properties (chemical properties) and root exudates (metabolites) had the greatest influence on the bacterial community of salt marsh, especially for abundant and moderate taxa. Our results provided novel insights into the biogeography of halophyte microbiome in coastal wetlands and can be beneficial for policymakers in decision-making on the management of coastal wetlands.

Pollution in river tributaries restricts the water quality of ecological water replenishment in the Baiyangdian watershed, China.

Natural rivers often have complex water network structures, and the continuous water inflow from tributaries may have crucial impacts on the water quality of ecological water replenishment in the mainstream. This study selected two main inflow rivers of the largest lake in Hebei Province (Baiyangdian), the Fu River and Baigou River, to explore the influence of tributaries on the quality changes of ecological replenishment water in the mainstreams. In December 2020 and 2021, water samples were collected along the two river routes, and eutrophic parameters and heavy metals were determined. The results showed that the tributaries of the Fu River were all severely polluted. With the inflows of the tributaries, the comprehensive pollution index of eutrophication greatly increased along the replenished water route of the Fu River, and the replenished water in the lower reaches of the Fu River mainstream was mostly considered moderate to heavy pollution. Whereas, because the Baigou River's tributaries were only moderately polluted, the water quality in the Baigou River's replenished water was mostly better than moderate pollution. Due to the slight pollution of heavy metals in the tributaries, the replenished water in both the Fu and Baigou Rivers did not show any impact from heavy metal pollution. Correlation and principal component analysis indicated that the main sources of serious eutrophic pollution in the tributaries of the Fu and Baigou Rivers were related to domestic sewage, industrial wastewater, plant decay, and sediment release. This non-point source pollution then caused the decline in the quality of the replenished water in the mainstreams. This study exposed a long-standing but neglected problem in ecological water replenishment and provided a scientific foundation for conducting better water management to improve the inland water environment.

A Rapid Identification Technique of Moving Loads Based on MobileNetV2 and Transfer Learning

Rapid and accurate identification of moving load is crucial for bridge operation management and early warning of overload events. However, it is hard to obtain them rapidly via traditional machine learning methods, due to their massive model parameters and complex network structure. To this end, this paper proposes a novel method to perform moving loads identification using MobileNetV2 and transfer learning. Specifically, the dynamic responses of a vehicle–bridge interaction system are firstly transformed into a two-dimensional time-frequency image by continuous wavelet transform to construct the database. Secondly, a pre-trained MobileNetV2 model based on ImageNet is transferred to the moving load identification task by transfer learning strategy for describing the mapping relationship between structural response and these specified moving loads. Then, load identification can be performed through inputting bridge responses into the established relationship. Finally, the effectiveness of the method is verified by numerical simulation. The results show that it can accurately identify the vehicle weight, vehicle speed information, and presents excellent strong robustness. In addition, MobileNetV2 has faster identification speed and requires less computational resources than several traditional deep convolutional neural network models in moving load identification, which can provide a novel idea for the rapid identification of moving loads.

Key Points-in-Time Identification of Gold Futures Market: A Complex Network Approach

Important nodes can determine the internal structure of complex networks and reveal the internal relationships of real-world systems, and identifying key nodes in complex networks is one of the important research areas of complex network science. As the king of commodities, changes in the price of gold significantly impact the economic development of various countries. Especially in the early stages of the outbreak of war between Russia and Ukraine, the price of gold futures has been greatly impacted, and the systemic risks are gradually spreading. In this paper, a gold future price series is mapped into a visibility graph (VG), the characteristics of the gold price time series and key points-in-time, have been explored from the perspective of complex network. First, according to the data structure characteristics of gold futures, this paper converts the closing prices of gold futures of the New York Mercantile Exchange into a complex network through the VG model. Then, by using the complex network model to further delve into the price of gold futures, it is found that the degree distribution of the gold futures network follows a power-law distribution, and has obvious scale-free characteristics. Finally, this paper uses the visual network node shrinking algorithm and the technique for order preference by similarity to ideal solution (TOPSIS) analysis method to identify the key nodes of the gold futures visual map to find the key time nodes in the timeline of gold futures market. Analysis of the key time nodes of this market by four methods reveals that the repetition rate of the key time nodes in the methods’ top 10 ranking is as high as 82.5%, indicating that the results obtained in this paper are robust. This study introduces a new model to describe the characteristics of gold futures price series, one which can find key time nodes in gold futures prices and provide potential help for predicting gold futures prices.

Single-Cell Labeling Strategies to Dissect Neuronal Structures and Local Functions.

The brain network consists of ten billion neurons and is the most complex structure in the universe. Understanding the structure of complex brain networks and neuronal functions is one of the main goals of modern neuroscience. Since the seminal invention of Golgi staining, single-cell labeling methods have been among the most potent approaches for dissecting neuronal structures and neural circuits. Furthermore, the development of sparse single-cell transgenic methods has enabled single-cell gene knockout studies to examine the local functions of various genes in neural circuits and synapses. Here, we review non-transgenic single-cell labeling methods and recent advances in transgenic strategies for sparse single neuronal labeling. These methods and strategies will fundamentally contribute to the understanding of brain structure and function.

Microbial network signatures of early colonizers in infants with eczema.

In this longitudinal cohort study, our results demonstrated that there are rhythmic changes in gut microbial network signatures in early life, and healthy infants adopt more complex and stable network structure in their gut microbiota than that of the infants with eczema.

Estimating Mixed Memberships in Directed Networks by Spectral Clustering

Community detection is an important and powerful way to understand the latent structure of complex networks in social network analysis. This paper considers the problem of estimating community memberships of nodes in a directed network, where a node may belong to multiple communities. For such a directed network, existing models either assume that each node belongs solely to one community or ignore variation in node degree. Here, a directed degree corrected mixed membership (DiDCMM) model is proposed by considering degree heterogeneity. An efficient spectral clustering algorithm with a theoretical guarantee of consistent estimation is designed to fit DiDCMM. We apply our algorithm to a small scale of computer-generated directed networks and several real-world directed networks.

Exploring the expressiveness of abstract metabolic networks.

Metabolism is characterised by chemical reactions linked to each other, creating a complex network structure. The whole metabolic network is divided into pathways of chemical reactions, such that every pathway is a metabolic function. A simplified representation of metabolism, which we call an abstract metabolic network, is a graph in which metabolic pathways are nodes and there is an edge between two nodes if their corresponding pathways share one or more compounds. The abstract metabolic network of a given organism results in a small network that requires low computational power to be analysed and makes it a suitable model to perform a large-scale comparison of organisms' metabolism. To explore the potentials and limits of such a basic representation, we considered a comprehensive set of KEGG organisms, represented through their abstract metabolic network. We performed pairwise comparisons using graph kernel methods and analyse the results through exploratory data analysis and machine learning techniques. The results show that abstract metabolic networks discriminate macro evolutionary events, indicating that they are expressive enough to capture key steps in metabolism evolution.

Model-independent embedding of directed networks into Euclidean and hyperbolic spaces

The arrangement of network nodes in hyperbolic spaces has become a widely studied problem, motivated by numerous results suggesting the existence of hidden metric spaces behind the structure of complex networks. Although several methods have already been developed for the hyperbolic embedding of undirected networks, approaches able to deal with directed networks are still in their infancy. Here, we present a framework based on the dimension reduction of proximity matrices reflecting the network topology, coupled with a general conversion method transforming Euclidean node coordinates into hyperbolic ones even for directed networks. While proposing a measure of proximity based on the shortest path length, we also incorporate an earlier Euclidean embedding method in our pipeline, demonstrating the widespread applicability of our Euclidean-hyperbolic conversion. Besides, we introduce a dimension reduction technique that maps the nodes directly into the hyperbolic space of any number of dimensions with the aim of reproducing a distance matrix measured on the given (un)directed network. According to various commonly used quality scores, our methods are capable of producing high-quality embeddings for several real networks.

OptiDepthNet: A Real-Time Unsupervised Monocular Depth Estimation Network

With the development of deep learning, the network architectures and algorithm accuracy applied to monocular depth estimation have been greatly improved. However, these complex network structures can be very difficult to realize real-time processing on embedded platforms. Consequently, this study proposed a lightweight encoding and decoding structure based on the U-Net model. The depthwise separable convolution was introduced into the encoder and decoder to optimize the network structure, further reduce the computational complexity, and improve the running speed, the implementation algorithm being more suitable for embedded platforms. When the accuracy of similar depth images was achieved, the network parameters could be reduced by up to eight times, and the running speed could be more than doubled. The research showed the proposed method to be very effective, having a certain reference value in monocular depth estimation algorithms running on embedded platforms.