Structure Of Large Networks Research Articles

Global stability for McKean–Vlasov equations on large networks

Abstract We investigate the mean-field dynamics of stochastic McKean differential equations with heterogeneous particle interactions described by large network structures. To express a wide range of graphs, from dense to sparse structures, we incorporate the recently developed graph limit theory of graphops into the limiting McKean–Vlasov equations. Global stability of the splay steady state is proven via a generalised entropy method, leading to explicit graph structure-dependent decay rates. We highlight the robustness of the entropy approach by extending the results to the closely related Sakaguchi–Kuramoto model with intrinsic frequency distributions. We also present central examples of random graphs, such as power law graphs and the spherical graphop, and analyse the limitations of the applied methodology.

Modular stochastic configuration network with attention mechanism for soft measurement of water quality parameters in wastewater treatment processes

Due to the complex and multi-distributed nature of wastewater treatment process data, stochastic configuration networks (SCNs) face difficulties such as large network structure and complex construction process when dealing with data modeling tasks. To address these challenges, this paper proposes a modular stochastic configuration network with attention mechanism, which integrates the advantages of modular partitioning, attention mechanism, and SCN. The model employs a fuzzy mean clustering strategy to simplify and decompose the complex multi-case and multi-distribution nonlinear modeling task, thereby reducing the modeling complexity of a single model. Furthermore, the model utilizes a stochastic configuration algorithm to learn the subtasks and construct corresponding sub-networks. The output weights are then assigned to each sub-network based on the attention mechanism to obtain the model’s final output. The validity of the proposed method is validated through two benchmark experiments. Then the model is applied to measure effluent ammonia nitrogen. The results indicate that compared with the classical randomized and modularized network models, the proposed model exhibits certain advantages in terms of learning accuracy and generalization performance.

GAP: A group-based automatic pruning algorithm via convolution kernel fusion

In recent years, the deployment and operation of convolution neural networks on edge devices with limited computing capabilities have become increasingly challenging due to the large network structure and computational cost. Currently, the mainstream structured pruning algorithms mainly compress the network at the filter or layer level. However, these methods introduce too much human intervention with large granularities, which may lead to unpredictable performance after compression. In this paper, we propose a group-based automatic pruning algorithm(GAP) via kernel fusion to automatically search for the optimal pruning structure in a more fine-grained manner. Specifically, we first adopt a novel nonlinear dimensionality reduction clustering algorithm to divide the filters of each convolution layer into groups of equal size. Afterwards, we encode the mutual distribution similarity of the kernels within each group, and its KL divergence is employed as an importance indicator to determine the retained kernel groups through weighted fusion. Subsequently, we introduce an intelligent searching module that automatically explore and optimize the pruned structure of each layer. Finally, the pruned filters are permutated to form a dense group convolution and fine-tuned. Sufficient experiments show that, on two image classification datasets, for five advanced CNN models, our GAP algorithm outperforms most extant SOTA schemes, reduces artificial intervention, and enables efficient end-to-end training of compact models.

Data-driven stock forecasting models based on neural networks: A review

As a core branch of financial forecasting, stock forecasting plays a crucial role for financial analysts, investors, and policymakers in managing risks and optimizing investment strategies, significantly enhancing the efficiency and effectiveness of economic decision-making. With the rapid development of information technology and computer science, data-driven neural network technologies have increasingly become the mainstream method for stock forecasting. Although recent review studies have provided a basic introduction to deep learning methods, they still lack detailed discussion on network architecture design and innovative details. Additionally, the latest research on emerging large language models and neural network structures has yet to be included in existing review literature. In light of this, this paper comprehensively reviews the literature on data-driven neural networks in the field of stock forecasting from 2015 to 2023, discussing various classic and innovative neural network structures, including Recurrent Neural Networks (RNNs), Convolutional Neural Networks (CNNs), Transformers, Graph Neural Networks (GNNs), Generative Adversarial Networks (GANs), and Large Language Models (LLMs). It analyzes the application and achievements of these models in stock market forecasting. Moreover, the article also outlines the commonly used datasets and various evaluation metrics in the field of stock forecasting, further exploring unresolved issues and potential future research directions, aiming to provide clear guidance and reference for researchers in stock forecasting.

Efficient evaporation of seawater desalination by novel double-interface evaporator: Photothermal conversion, water transfer and salt-resistant

Interfacial solar-driven water evaporation emerged as an established eco-friendly method for seawater desalination to combat water scarcity, yet the low water evaporation rate and cost-effectiveness limit the practicable application. Herein, we presented a double-interface evaporator with a bilayer structure (BCCu@FC). The upper layer was constructed using carbonized copper-doped biochar (BCCu), which exhibited broadband sunlight absorption and effective photothermal conversion capability for the localized surface plasmon resonance (LSPR) effect. Meanwhile, the lower layer was constructed by felt fabric (FC), which featured a large spatial network structure, guaranteeing sufficient water transportation. Therefore, BCCu@FC exhibited a high photothermal conversion efficiency of 87.05% with a 1.571 kg m-2 h-1 water evaporation rate under 1 sun illumination. Moreover, BCCu@FC further exhibited high stability and salt resistance, showing no significant salt crust formation after continuous operation for 24 h under 3.5 wt% brine. Especially, BCCu@FC could stably work under 21 wt% brine, and the salt crust re-dissolved within 4 h after the cessation of illumination. This facile and universal strategy was anticipated to pave the way for future designs and fabrication of diverse, more efficient, and cost-effective photothermal conversion devices.

Neural Network with Local Converging Input for Unstructured-Grid Computational Fluid Dynamics

In recent years, surrogate models based on deep neural networks have been widely used to solve partial differential equations for fluid flow physics. This kind of model focuses on global interpolation of the training data and thus requires a large network structure. The process is both time consuming and computationally costly. In the present study, we develop a neural network with local converging input (NNLCI) for high-fidelity prediction using unstructured data. The framework uses the local domain of dependence with converging coarse solutions as input, thereby greatly reducing computational resource and training time. As a validation case, the NNLCI method is applied to study two-dimensional inviscid supersonic flows in channels with bumps. Different bump geometries and locations are examined to benchmark the effectiveness and versatility of this new approach. The NNLCI method can accurately and efficiently capture the structure and dynamics of the entire flowfield, including regions with shock discontinuities. For a new bump configuration, the method can perform prediction with only one neural network, eliminating the need for repeated training of multiple networks for different geometries. A saving of computing wall time is achieved by several orders of magnitude against the high-fidelity simulation with the same level of accuracy. The demand on training data is modest, and the training data can be allocated sparsely. These features are especially advantageous compared with conventional global-to-global deep learning methods and physics-informed methods.

블록모델링을 활용한 한국자동차산업 공급네트워크 구조 분석

This study empirically analyzes the supply network of the Korean automotive industry through CONCOR, a structural equivalence analysis, and blockmodeling. Using equivalence analysis, it is possible to classify nodes based on occupying the same position in the network and identify the relationship pattern between them, even if it is not based on direct connections between nodes. The results of the study are as follows First, the supply network of the Korean automobile industry was divided into six blocks, and the six domestic automakers belonged to different blocks. Second, we found that blocks 4 and 5 were significantly larger than blocks 1 and 2 in terms of attributes such as sales and total assets. Finally, the relational characteristics between the blocks suggest that block 5 plays a pivotal role in the supply network of the Korean automobile industry. The theoretical significance of this study is that it is the first to apply the concept of structural equivalence to the field of SCM research, providing a new way to systematically understand the structure of large and complex supply networks. In addition, from a practical perspective, it can help managers to view and understand the entire supply network from a macro perspective.

Ammonia-assisted Ni particle preferential deposition in Ni-Fe pyrophosphates on iron foam to improve the catalytic performance for overall water splitting

Designing efficient and cost-effective electrocatalysts for overall water splitting remains a major challenge in hydrogen production. Herein, ammonia was introduced to pyrophosphate chelating solution assisted Ni particles preferential plating on porous Fe substrate to form coral-like Ni/NiFe-Pyro electrode. The pyrophosphate with multiple complex sites can couple with nickel and iron ions to form an integrated network structure, which also consists of metallic nickel due to the introduction of ammonia. The large network structure in Ni/NiFe-Pyro significantly enhances the synergistic effect between nickel and iron and then improves the electrocatalytic performance. As a result, the coral-like Ni/NiFe-Pyro@IF exhibits good electrocatalytic activity and stability for oxygen evolution reaction (OER) and hydrogen evolution reaction (HER). The electrolyzer assembled with Ni/NiFe-Pyro@IF as cathode and anode just needs a low water-splitting voltage of 1.54 V to obtain the current density of 10 mA cm−2. Meanwhile, the stability test of Ni/NiFe-Pyro@IF is performed at the current densities ranging from 10 to 400 mA cm−2 for 50 h without any significant decay, indicating robust catalytic stability for overall water splitting. This strategy for synthesizing metal/metal pyrophosphate composites may provide a new avenue for future studies of efficient bifunctional electrocatalysts.

Highly salt-resistant 3D melamine sponge-polypyrrole composites for efficient solar interfacial evaporation

Solar desalination has received great attention as an environmentally friendly and cost-effective way to desalinate seawater. In this paper, a melamine sponge composite polypyrrole (MSPPy) evaporator was prepared by in-situ polymerization of Pyrrole (Py) on a melamine sponge (MS). PPy has an efficient light absorption ability. The large aperture network structure of MS can realize multiple reflections of light and greatly increase the absorption of light. Their synergistic effect makes the light absorption efficiency of the MSPPy evaporator reach 96.97%. Under one sun, the average evaporation rate of 1.886 kg⋅m−2⋅h−1 and energy conversion efficiency of 96.8% were achieved during 10 h. Moreover, the evaporator has an edge-priority crystallization function, which enables it to work for a long time in a high-salt environment. It works continuously for 12 h in simulated seawater with a salinity of 20% without further accumulation of salt crystals on its surface. More important, it can rapidly diffuse accumulated salt into the bulk water at night. This study provides a new idea for the application of solar steam generation technology in high-salt wastewater treatment.

Graph Autoencoders for Embedding Learning in Brain Networks and Major Depressive Disorder Identification.

Brain functional connectivity (FC) networks inferred from functional magnetic resonance imaging (fMRI) have shown altered or aberrant brain functional connectome in various neuropsychiatric disorders. Recent application of deep neural networks to connectome-based classification mostly relies on traditional convolutional neural networks (CNNs) using input FCs on a regular Euclidean grid to learn spatial maps of brain networks neglecting the topological information of the brain networks, leading to potentially sub-optimal performance in brain disorder identification. We propose a novel graph deep learning framework that leverages non-Euclidean information inherent in the graph structure for classifying brain networks in major depressive disorder (MDD). We introduce a novel graph autoencoder (GAE) architecture, built upon graph convolutional networks (GCNs), to embed the topological structure and node content of large fMRI networks into low-dimensional representations. For constructing the brain networks, we employ the Ledoit-Wolf (LDW) shrinkage method to efficiently estimate high-dimensional FC metrics from fMRI data. We explore both supervised and unsupervised techniques for graph embedding learning. The resulting embeddings serve as feature inputs for a deep fully-connected neural network (FCNN) to distinguish MDD from healthy controls (HCs). Evaluating our model on resting-state fMRI MDD dataset, we observe that the GAE-FCNN outperforms several state-of-the-art methods for brain connectome classification, achieving the highest accuracy when using LDW-FC edges as node features. The graph embeddings of fMRI FC networks also reveal significant group differences between MDD and HCs. Our framework demonstrates the feasibility of learning graph embeddings from brain networks, providing valuable discriminative information for diagnosing brain disorders.

DNA digestion and formation of DNA-network structures with Holliday junction-resolving enzyme Hjc_15-6 in conjunction with polymerase reactions

The recently identified novel Holliday junction-resolving enzyme, termed Hjc_15-6, activity investigation results imply DNA cleavage by Hjc_15-6 in a manner that potentially enhances the molecular self-assembly that may be exploited for creating DNA-networks and nanostructures. The study also demonstrates Pwo DNA polymerase acting in combination with Hjc_15-6 capability to produce large amounts of DNA that transforms into large DNA-network structures even without DNA template and primers. Furthermore, it is demonstrated that Hjc_15-6 prefers Holliday junction oligonucleotides as compared to Y-shaped oligonucleotides as well as efficiently cleaves typical branched products from isothermal DNA amplification of both linear and circular DNA templates amplified by phi29-like DNA polymerase. The assembly of large DNA network structures was observed in real time, by transmission electron microscopy, on negative stained grids that were freshly prepared, and also on the same grids after incubation for 4 days under constant cooling. Hence, Hjc_15-6 is a promising molecular tool for efficient production of various DNA origamis that may be implemented for a wide range of applications such as within medical biomaterials, catalytic materials, molecular devices and biosensors.

Highly Breathable, Stretchable, and Tailorable TPU Foam for Flexible Gas Sensors.

Continuous real-time monitoring of air quality is of great significance in the realms of environmental monitoring, personal safety, and healthcare. Recently, flexible gas sensors have gained great popularity for their potential to be integrated into various smart wearable electronics and display devices. However, the development of gas sensors with superior sensitivity, breathability, and stretchability remains a challenge. Here, a new high porosity thermoplastic polyurethane (HP-TPU) foam was reported for gas sensors, which exhibited large three-dimensional network structures and excellent mechanical properties. The HP-TPU foam was achieved by using a simple steam-induced method, which was suitable for mass production. The unique structure endowed this foam with 77.5% porosity, 260% strain ability, and 0.45 MPa Young's modulus, which improved 35, 31, and 80%, respectively, compared to previously reported traditional TPU foam (T-TPU) prepared by the drying method. In addition, the foam presented high gas permeability (312 g/m-2, 24 h) and excellent stability, and it remained undamaged even after 2000 cycles at 70% strain. The sensing material was coated on a PET flexible interdigital electrode and sandwiched between two HP-TPU foam layers for a gas sensitivity test. Due to the easy diffusion of gas between the pores and contact with the sensing materials, the HP-TPU foam exhibited a significant reduction of 85% in average response time and 46% in average recovery time, compared to the T-TPU foam. A wearable sensing device, comprising sensing, data processing, and wireless transmission modules, was successfully developed to enable outdoor testing and achieved a detection range at the ppb level. Finally, the cytotoxicity test results confirmed that this flexible gas sensor did not harm human health. These results proved that this HP-TPU foam was an ideal matrix for the flexible gas sensor, exhibiting great application potential in the fields of seamless human-machine integration.

Synergistic effect of hydrophobic modified Welan gum and sodium alcohol ether sulphate in saline solution

Surfactants are often used to change the structure of natural polysaccharide polymers and improve their properties for industrial applications. However, implementing surfactants to boost the Welan gum has yet to be studied. This study prepared the compound system HWG-A by Sodium alcohol ether sulphate (AES) with hydrophobic modified Welan gum (HWG). Compared with HWG, the critical association concentration of HWG-A they have decreased from 3.8 × 103 mg L-1 to 2.8 × 103 mg L-1. The viscosity of 2 × 103 mg L-1 HWG-A remained at 244 mPa s when the salinity was 96,670 mg L-1. The rheological test showed that HWG-A still maintained the characteristics of pseudoplastic fluid under different polysaccharide concentrations, pH and salinity. The viscoelastic test showed that the elasticity of HWG-A was interdependent with the concentration of polysaccharides. When HWG-A was higher than the critical association concentration, the elasticity dominated the whole system. The temperature resistance test showed that the viscosity of HWG-A was still much higher than that of Welan gum and HWG when the salinity was 19,334 mg L-1 at 140 ℃ and lower than the critical association concentration, and the viscosity remained above 350 mPa s. SEM showed that HWG-A presented a regular large network structure when the concentration was lower than its critical association concentration. Otherwise, HWG-A presented a gel shape. Our study promotes the industrial potential of Welan gum in such as Enhanced Oil Recovery (EOR) process. It suggests that the interaction between surfactants and natural polysaccharides is essential in determining the functional properties of the fluid system.

YOLOv6-ESG: A Lightweight Seafood Detection Method

The rapid development of convolutional neural networks has significant implications for automated underwater fishing operations. Among these, object detection algorithms based on underwater robots have become a hot topic in both academic and applied research. Due to the complexity of underwater imaging environments, many studies have employed large network structures to enhance the model’s detection accuracy. However, such models contain many parameters and consume substantial memory, making them less suitable for small devices with limited memory and computing capabilities. To address these issues, a YOLOv6-based lightweight underwater object detection model, YOLOv6-ESG, is proposed to detect seafood, such as echinus, holothurian, starfish, and scallop. First, a more lightweight backbone network is designed by rebuilding the EfficientNetv2 with a lightweight ODConv module to reduce the number of parameters and floating-point operations. Then, this study improves the neck layer using lightweight GSConv and VoVGSCSP modules to enhance the network’s ability to detect small objects. Meanwhile, to improve the detection accuracy of small underwater objects with poor image quality and low resolution, the SPD-Conv module is also integrated into the two parts of the model. Finally, the Adan optimizer is utilized to speed up model convergence and further improve detection accuracy. To address the issue of interference objects in the URPC2022 dataset, data cleaning has been conducted, followed by experiments on the cleaned dataset. The proposed model achieves 86.6% mAP while the detection speed (batch size = 1) reaches 50.66 FPS. Compared to YOLOv6, the proposed model not only maintains almost the same level of detection accuracy but also achieves faster detection speed. Moreover, the number of parameters and floating-point operations reaches the minimum levels, with reductions of 75.44% and 79.64%, respectively. These results indicate the feasibility of the proposed model in the application of underwater detection tasks.

Characteristics of dough prepared with coconut oil

AbstractBackground and objectivesThis study investigated the use of coconut oil as an alternative to butter and shortening in bread making by analyzing the dough characteristics for feasibility.FindingsIn the fermentation test, the pH of the dough containing oils and fats was reduced to <5.1 of the threshold, which deteriorated the extensibility. The pH further decreased with an increased CO2 emission rate from the dough. In the compression test, the stress of the dough containing oils and fats was lower than that without it. However, the stress on gluten‐containing oils and fats was significantly high and differed based on the type of oils and fats. A large network structure of gluten with distributed starch granules was observed using scanning electron microscopy in the dough containing oils and fats. Numerous starch granules were uniformly distributed in the spaces of the gluten network structure, making the dough softer.ConclusionsIt was recognized that the characteristics of the dough containing coconut oil were similar to that of the dough containing butter or shortening.Significance and noveltyThis study suggested that coconut oil can be used as an alternative oil and fat to butter or shortening for bread making.

The public sphere as a dynamic network

Abstract This article proposes to conceptualize the public sphere as a dynamic network of actors and contents that are linked with each other by communicative actions. This perspective allows us to theoretically derive and empirically describe the entire range of small to large network structures and their evolution over time. First, we will define the elements of these networks, which include the actors, content, communicative actions, and content relations. Based on these entities, four communicative roles (producer, recipient, curator, isolate) will be distinguished. Second, we will summarize how these actors perceive the communicative situation and how they select from behavioral options. Third, we will show how this combines with the network dynamics and outcomes that are discussed in the different lines of research. This provides not only the basis for understanding the link between the communicative actions on the micro-level and macro-level structures, but also new avenues for normative discussions.

Optimization of Mopan Persimmon Wine Fermentation with Pectinase and Analysis of Its Mechanism of Action

Due to the high sugar content of Mopan persimmon, which has an annual output of more than 0.5 million tons in China, it can be processed to make fruit wine. In this study, a strain of yeast screened from different persimmon samples was used for persimmon wine fermentation. The optimal conditions of persimmon wine fermentation were determined through single-factor experiments as follows: Yeast addition of 0.08 g/kg; a fermentation temperature of 28 °C; sucrose addition of 18%; and pectinase addition of 0.01%. Under these conditions, the alcohol content of persimmon wine reached 12.9%. The addition of pectinase during persimmon wine fermentation was found to decompose pectin at high speed, reduce the viscosity of the fermentation liquid, increase the dissolved oxygen content in the fermentation liquid, promote the growth and reproduction of yeast, and effectively convert the sugars into alcohol. After fermentation, alcohol, residual sugars, and total phenolic content with or without pectinase treatment were 12.9 and 4.4%, 2.2 and 13.4 g/L, and 738.7 and 302 µg/mL, respectively. Scanning electron microscopy (SEM) results showed that compared with the mash without pectinase treatment, the mash with pectinase had a larger network structure and more pores and yeasts.

The flexible EDAN-based covalent organic frameworks constructed by flexible knots and 4,4′-ethylenedianiline as a linker for adsorping and fluorescence sensing iodine

In this paper, 4,4′-ethylenedianiline as a linker, flexible 2,4,6-tris(4-formylphenoxy)- 1,3,5-triazine (TPT−3−CHO) and hexa(4-formyl-phenoxy)cyclotriphosphazene (NOP−6−CHO) as knots were selected to construct the new flexible EDAN-based COFs (TEDAN and HEDAN) through Schiff base reactions of aldehyde and amine. The as-prepared flexible EDAN-based COFs have high crystallinity, sizable surface area, and thermal stability. With porous properties, abundant electron-rich heteroatoms as well as large π-conjugated network structures that can provide effective binding sites, TEDAN and HEDAN exhibit good reversible vapor-trapped iodine with the adsorption capacities of 4.64 and 2.25 g g−1, respectively. The aromatic networks of nitrogen-rich units in TEDAN provide many binding sites with guest iodine molecules than that of HEDAN, resulting in higher equilibrium uptake capacity of iodine for TEDAN. In addition, the flexible EDAN-based COFs show “turn-off” detection of non-radioactive iodine in the dispersion state with quenched constants (KSV) of 5.59 × 104 and 2.35 × 104 L mol−1. The fluorescence quenching of the flexible TEDAN-based COFs can be attributed to both the photo-induced electron transfer mechanism and the energy transfer transfer mechanism. The sensitivity of HEDAN is higher than that of TEDAN.

Identification of an influence network using ensemble-based filtering for Hawkes processes driven by count data

Many networks have event-driven dynamics (such as communication, social media and criminal networks), where the mean rate of the events occurring at a node in the network changes according to the occurrence of other events in the network. In particular, events associated with a node of the network could increase the rate of events at other nodes, depending on their influence relationship. Thus, it is of interest to use temporal data to uncover the directional, time-dependent, influence structure of a given network while also quantifying uncertainty even when knowledge of a physical network is lacking. Typically, methods for inferring the influence structure in networks require knowledge of a physical network or are only able to infer small network structures. In this paper, we model event-driven dynamics on a network by a multidimensional Hawkes process. We then develop a novel ensemble-based filtering approach for a time-series of count data (i.e., data that provides the number of events per unit time for each node in the network) that not only tracks the influence network structure over time but also approximates the uncertainty via ensemble spread. The method overcomes several deficiencies in existing methods such as existing methods for inferring multidimensional Hawkes processes are too slow to be practical for any network over ∼50 nodes, can only deal with timestamp data (i.e. data on just when events occur not the number of events at each node), and that we do not need a physical network to start with. Our method is massively parallelizable, allowing for its use to infer the influence structure of large networks (∼10,000 nodes). We demonstrate our method for large networks using both synthetic and real-world email communication data.

A super-resolution network using channel attention retention for pathology images.

Image super-resolution (SR) significantly improves the quality of low-resolution images, and is widely used for image reconstruction in various fields. Although the existing SR methods have achieved distinguished results in objective metrics, most methods focus on real-world images and employ large and complex network structures, which are inefficient for medical diagnosis scenarios. To address the aforementioned issues, the distinction between pathology images and real-world images was investigated, and an SR Network with a wider and deeper attention module called Channel Attention Retention is proposed to obtain SR images with enhanced high-frequency features. This network captures contextual information within and across blocks via residual skips and balances the performance and efficiency by controlling the number of blocks. Meanwhile, a new linear loss was introduced to optimize the network. To evaluate the work and compare multiple SR works, a benchmark dataset bcSR was created, which forces a model training on wider and more critical regions. The results show that the proposed model outperforms state-of-the-art methods in both performance and efficiency, and the newly created dataset significantly improves the reconstruction quality of all compared models. Moreover, image classification experiments demonstrate that the suggested network improves the performance of downstream tasks in medical diagnosis scenarios. The proposed network and dataset provide effective priors for the SR task of pathology images, which significantly improves the diagnosis of relevant medical staff. The source code and the dataset are available on https://github.com/MoyangSensei/CARN-Pytorch.