Network Of Processes Research Articles

Classification of Spatial Temporal Patterns Based on Neuromorphic Networks

This work is devoted to the problems of developing neuromorphic classifiers of spatiotemporal patterns, as well as their application in neurointerfaces. Classifiers of spatiotemporal patterns based on neural networks, support vector machines, deep neural networks, and Riemannian geometry are considered. A comparative study of these classifiers is carried out in the plane of the accuracy of multiclass recognition of electroencephalographic signals showing time-dependent bioelectrical activity in different areas of the brain during the imagination of different movements. It is shown that such classifiers can provide an accuracy of 60-80% when recognizing from two to four classes of imaginary movements. A new type of classifier based on a neuromorphic network, based on the biosimilar neurons built on the Izhikevich model, is proposed. The network processes input spike sequences and generates pulse streams of different frequencies at the outputs. The network is trained using the Supervised STDP algorithm based on labeled information containing examples of the correct recognition of the required pattern classes. The recognized pattern class is determined by the maximum frequency of the output sequence. The neuromorphic classifier showed an average classification accuracy of 90% for 4 classes of imaginary commands and a maximum of 95%. By modeling the robot control task in the virtual environment it is shown that such accuracy is sufficient for the effective use of the classifier as part of a non-invasive brain-computer interface for non-contact control of robotic devices.

Theory guided Lagrange programming neural network for subsurface flow problems

A deep learning model can perform efficient uncertainty quantification (UQ) for reservoir flow with uncertain model parameters, but usually requires large amounts of training data to ensure accuracy. However, the cost of obtaining large amounts of data is prohibitive, and the performance will deteriorate if sufficient training data is lacking. Alternatively, more interpretable neural networks with embedded physical laws have recently been used to solve partial differential equations as well as to solve UQ problems. This approach has received a lot of attention due to its low data volume requirements and its adherence to the laws of physics during the training process. In this paper, we propose a theory-guided framework based on a bilevel programming model with hard constraints to embed physical meaning in the model. Theory guided Lagrange programming neural network (TGLPNN) combines the method of Lagrange programming neural network approach where physical laws such as stochastic partial differential equations and boundary conditions are incorporated into the training process of a convolutional neural network. At the same time, the upper-level variables are iteratively optimized. The method based on Lagrange programming neural network inherently embeds physical laws in the network. Practical applications have shown that TGLPNN can provide higher prediction accuracy compared to state-of-the-art physics-driven methods and improved efficiency compared to numerical methods.

Meta learning based residual network for industrial production quality prediction with limited data

Due to the challenge of collecting a substantial amount of production-quality data in real-world industrial settings, the implementation of production quality prediction models based on deep learning is not effective. To achieve the goal of predicting production quality with limited data and address the issue of model degradation in the training process of deep learning networks, we propose Meta-Learning based on Residual Network (MLRN) models for production quality prediction with limited data. Firstly, the MLRN model is trained on a variety of learning tasks to acquire knowledge for predicting production quality. Furthermore, to obtain more features with limited data and avoid the issues of gradient disappearing or exploding in deep network training, the enhanced residual network with the effective channel attention (ECA) mechanism is chosen as the basic network structure of MLRN. Additionally, a multi-batch and multi-task data input approach is implemented to prevent overfitting. Finally, the availability of the MLRN model is demonstrated by comparing it with other models using both numerical and graphical datasets.

Skip connection information enhancement network for retinal vessel segmentation.

Many major diseases of the retina often show symptoms of lesions in the fundus of the eye. The extraction of blood vessels from retinal fundus images is essential to assist doctors. Some of the existing methods do not fully extract the detailed features of retinal images or lose some information, making it difficult to accurately segment capillaries located at the edges of the images. In this paper, we propose a multi-scale retinal vessel segmentation network (SCIE_Net) based on skip connection information enhancement. Firstly, the network processes retinal images at multiple scales to achieve network capture of features at different scales. Secondly, the feature aggregation module is proposed to aggregate the rich information of the shallow network. Further, the skip connection information enhancement module is proposed to take into account the detailed features of the shallow layer and the advanced features of the deeper network to avoid the problem of incomplete information interaction between the layers of the network. Finally, SCIE_Net achieves better vessel segmentation performance and results on the publicly available retinal image standard datasets DRIVE, CHASE_DB1, and STARE.

PerfTop: Towards performance prediction of distributed learning over general topology

Distributed learning with multiple GPUs has been widely adopted to accelerate the training process of large-scale deep neural networks. However, misconfiguration of the GPU clusters with various communication primitives and topologies could potentially diminish the gains in parallel computation and lead to significant degradation in training efficiency. Predicting the performance of distributed learning enables service providers to identify potential bottlenecks beforehand. In this work, we propose a Performance prediction framework over General Topologies, called PerfTop, for accurate estimation of per-iteration execution time. The main strategy is to integrate computation time prediction with an analytical model to map the nonlinearity in communication and fine-grained computation-communication patterns. This enables accurate prediction of a variety of neural network models over general topologies, such as tree, hierarchical, and exponential. Our extensive experiments show that PerfTop outperforms existing methods in estimating both computation and communication time, particularly for communication, surpassing the existing methods by over 45%. Meanwhile, it achieves an accuracy of above 85% in predicting the execution time over general topologies compared to simple topologies such as star and ring from the previous works.

Total Quality Management in the Service Sector: A Case Study in a Water Supply Company

Objective: Analyze the benefits of adopting total quality in a water supply company, aiming to improve service quality and customer satisfaction. Theoretical Framework: A literature review was conducted as a basis for understanding the key concepts of TQM, the dimensions of service quality, and the essential elements for the successful implementation of TQM in the context of the analyzed company. Method: The research is classified as a case study with an applied qualitative approach. To achieve the proposed objectives, techniques such as documentary research, literature review, semi-structured interviews, and direct observation were used. Results: The analysis of the results demonstrated that the water supply company faces significant challenges regarding service quality. From the TQM approach, opportunities for improvement were identified in different areas of the organization, both in the reliability of the services' outcomes, including the water treatment process and maintenance of the distribution network, and in customer service. Research Implications: The research suggests that the water supply company should adopt TQM as a strategy to improve service quality, promote customer satisfaction, address internal challenges, and seek constant improvement. Originality/Value: The study explores the application of TQM in a public service context, providing practical guidance and insights to improve service quality in a water supply company, with implications that may extend to similar organizations seeking to enhance their services and processes.

A Kalman filter approach for state estimation in weakly monitored active distribution networks

Distribution networks are becoming active to improve security of operation under high net load volatility. Yet, the deployment of active devices creates new challenges to the operation of the weakly monitored distribution networks. One such challenge is to improve state estimation accuracy. This paper proposes to improve the estimation process of active distribution networks by embedding the active devices dynamics into the state dynamic equations, instead of translating them into the static measurement models. This is carried out with the Kalman filter (KF) by making use of the dynamic models and the static measurement models we have proposed earlier for active distribution networks. The paper provides the main steps to implement the estimation framework and illustrates its advantages under slow-responsive distribution network interactions. Results obtained show that the proposed KF approach has better performance than the static counterparts for networks subjected to multiple control decisions — the case where the relevance information on dynamics surpasses the relevance of setpoint information on static models alone.

Fuzzy control for networked systems with mixed data missing under quantization based event-triggered scheme

This paper presents an innovative event-triggered fuzzy control approach for networked systems operating over unreliable network channels. Traditional complicated modeling and analysis process of network induced issues is replaced by a bran-new logarithmic quantization based event-triggered scheme method. The joint characterization of mixed data missing, arising from the event-triggered mechanism and unreliable network channels, becomes a challenging issue due to disrupted data sequences and variable relations. Consequently, this paper contributes a formulated auxiliary stochastic sequence approach to overcome the difficulty of dynamics modeling. With an formulated event driven controller, the modeling and implementation of the closed-loop systems are constructed and further simplified. Considering the difference and relation between the membership functions of plant and controller, we present a new fuzzy output feedback controller design method to obtain feasible result via uncertainty based method. Finally, we verify the validity of the designed control policy via a simulation example.

Programmable Retention Characteristics in MoS2-Based Atomristors for Neuromorphic and Reservoir Computing Systems.

In this study, we investigate the coexistence of short- and long-term memory effects owing to the programmable retention characteristics of a two-dimensional Au/MoS2/Au atomristor device and determine the impact of these effects on synaptic properties. This device is constructed using bilayer MoS2 in a crossbar structure. The presence of both short- and long-term memory characteristics is proposed by using a filament model within the bilayer transition-metal dichalcogenide. Short- and long-term properties are validated based on programmable multilevel retention tests. Moreover, we confirm various synaptic characteristics of the device, demonstrating its potential use as a synaptic device in a neuromorphic system. Excitatory postsynaptic current, paired-pulse facilitation, spike-rate-dependent plasticity, and spike-number-dependent plasticity synaptic applications are implemented by operating the device at a low-conductance level. Furthermore, long-term potentiation and depression exhibit symmetrical properties at high-conductance levels. Synaptic learning and forgetting characteristics are emulated using programmable retention properties and composite synaptic plasticity. The learning process of artificial neural networks is used to achieve high pattern recognition accuracy, thereby demonstrating the suitability of the use of the device in a neuromorphic system. Finally, the device is used as a physical reservoir with time-dependent inputs to realize reservoir computing by using short-term memory properties. Our study reveals that the proposed device can be applied in artificial intelligence-based computing applications by utilizing its programmable retention properties.

A knowledge distillation based cross-modal learning framework for the lithium-ion battery state of health estimation

The accurate prediction of a lithium-ion battery’s State of Health is of critical importance for efficient battery health management. Existing data-driven estimation methodologies grapple with issues such as high model complexity and a dearth of guidance from prior knowledge, which impose constraints on their efficacy. This work introduces a novel cross-modal distillation network for battery State of Health estimation, structured around a TransformerEncoder as the teacher network and a Convolutional Neural Network as the student network. Initially, the teacher model is pre-trained offline using State of Health degradation data to learn the degradation patterns. The directly measurable feature data (such as voltage, temperature, and current) is subsequently fed into the student network for online training and computation of a hard loss. the student network’s output is then directed into the pre-trained the teacher network to compute a soft loss, thereby offering prior knowledge of degradation laws and steering the optimization process of the student network. Rigorous experiments are conducted utilizing various datasets, with the outcomes validating the superior estimation accuracy and degradation rule adherence of the model. Notably, among five different models, this model demonstrates the best performance on almost all datasets, achieving an RMSE of 0.0097 and an MAE of 0.0065 on Cell1 of the Oxford dataset. Moreover, the model also demonstrates robust performance across different usage scenarios, inclusive of multi-battery estimation. Furthermore, this paper also introduces a fine tuning method for State of Health predictions only using the first half of the data. Comparative analysis with other models underscores the competitiveness of the proposed model, showcasing its potential for broader application.

Drifting towards multidimensional collaborative innovation: the co-evolution of collaboration networks and policy in China’s nuclear power industry

Purpose Nuclear power is a stable and reliable energy source that can improve energy structure while reducing carbon emissions, which is of great significance for environmental protection and combating climate change. As a unique industry, it is facing rare development opportunities in China and has broad market prospects. However, the characteristics of technical difficulty, loose organizational structure and uneven regional distribution limit the expansion of the nuclear power industry. This paper aims to a better understanding of the accumulation process for innovation capability from the perspective of network evolution and provides policy guidance for the market development of the nuclear power industry (NPI). Design/methodology/approach Methodologically, social network analysis is used to explore the co-evolution of multidimensional collaboration networks. First, the development and policy evolution of the NPI is introduced to divide the evolution periods. Then, the authors identify and analyze the core organizations, technologies and regions that promote nuclear power patent collaboration. Furthermore, three levels of collaboration networks based on organizations, technologies and regions are constructed to analyze the coevolution of patent networks in China’s NPI. Findings The results show that nuclear power enterprises always play the foremost role in the organizational collaboration network (OCN), and the dominance of foreign enterprises is replaced by Chinese state-owned enterprises in the third period. The technology hotspot has shifted from nuclear power plant construction to the control system. The regional collaboration network was initially formed in the coastal areas and gradually moved inland, with Guangdong and Beijing becoming the two cores of the network. The scale of three collaboration networks is still expanding but the speed has slowed down. Originality/value In response to the pain points of the NPI, this research focuses on multidimensional collaborative innovation, investigates the dynamic evolution process of collaborative innovation networks in China’s NPI and links policy evolution with network evolution creatively. The ultimate result not only helps nuclear power enterprises integrate innovative resources in complex environments but also promotes industrial upgrading and market development.

Tough and biodegradable C-lignin cross-linked polyvinyl alcohol supramolecular composite films with closed-looping recyclability

Catechyl lignin (C-lignin) is a unique natural biopolymer that is considered to be the best alternative to conventional lignin due to its linear homogeneous molecular structure. Based on the structural properties of C-lignin, the development of a new biodegradable plastic to replace traditional petroleum-based plastics is a sustainable route. However, the maximum filling fraction of lignin is required to achieve excellent biodegradability of lignin-based bio-plastics, yet resulting in decreased mechanical strength of composites. Herein, we report a novel strategy to fabricate a closed-loopling recyclability supramolecular composite films through casting and dry molding process of intermolecular covalent-noncovalent cross-linking networks with polyvinyl alcohol by the epoxidation of C-lignin. The intermolecular cross-linking mechanism analyses reveal that the resultant bio-material not only display improved mechanical properties (maximum tensile strength of 133.5 MPa and strain at break of 690.2 %), but also show UV-resistant, antimicrobial, thermally stable, and rapid biodegradable properties. Notably, the fabricated films demonstrate the characteristics of closed-loop recyclability, which paves a new way toward rational design of bioplastics using renewable lignin. These integrated properties make the lignin-based biocomposites a promising alternative to traditional plastics.

Variance of the Infection Number of Heterogeneous Malware Spread in Network

The Susceptible–Infected–Susceptible (SIS) model in complex networks is one of the critical models employed in the modeling of virus spread. The study of the heterogeneous SIS model with a non-homogeneous nodal infection rate in finite-size networks has attracted little attention. Investigating the statistical properties of heterogeneous SIS epidemic dynamics in finite networks is thus intriguing. In this paper, we focus on the measure of variability in the number of infected nodes for the heterogeneous SIS epidemic dynamics in finite-size bipartite graphs and star graphs. Specifically, we investigate the metastable-state variance of the number of infected nodes for the SIS epidemic process in finite-size bipartite graphs and star graphs with heterogeneous nodal infection rates. We employ an extended individual-based mean-field approximation to analyze the heterogeneous SIS epidemic process in finite-size bipartite networks and star graphs. We derive the approximation solutions of the variance of the infected number. We verify the proposed theory by simulations. The proposed theory has the potential to help us better understand the fluctuations of SIS models like epidemic dynamics with a non-homogeneous infection rate.

A Distributed Deployment Method for Wireless Sensor Networks Based on Multi Agent Systems

This paper proposes a distributed deployment method for wireless sensor networks based on multi-agent systems, which addresses the issues of poor self adaptive deployment capability and high deployment costs caused by numerous and diverse types of nodes in the deployment process of wireless sensor networks, leading to blind spots in coverage and redundant nodes. Due to the characteristics of intelligent perception, intelligent communication, and clustering of nodes in wireless sensor network, sensor nodes can be seen as sensing agents, and sensor networks can be seen as distributed multi-agent systems composed of numerous sensing agents. Through the sensing agents' perception of the surrounding environment and communication interaction, a unified deployment model and distributed deployment algorithm with multiple nodes (cluster head nodes) as leaders are established, achieving effective integration of autonomous and leadership mechanisms, distributed management, and centralized management. Taking the problem of sensor network coverage as an example, simulation experiments were conducted, including 12 sensor agents and 1 20 * 20 monitoring areas, to verify how the sensor agents perceive the environment and optimize the migration and deployment process under the coordination of cluster head nodes. The experimental results confirmed the effectiveness of the above models and algorithms. The research results provide a theoretical model and practical approach for the distributed deployment and scheduling of sensor nodes in wireless sensor networks, and can be extended to intelligent Internet of Things systems, such as urban intelligent agent systems, logistics intelligent agent systems, etc., to improve the intelligence of wireless sensor networks deployment and scheduling.

MTTLA‐DLW: Multi‐task TCN‐Bi‐LSTM transfer learning approach with dynamic loss weights based on feature correlations of the training samples for short‐term wind power prediction

AbstractWind power prediction for newly built wind farms is usually faced with the problem of no sufficient historical data. To efficiently extract the useful features from related wind farms, a novel transfer learning method based on temporal convolutional network (TCN)‐Bi‐long short‐term memory (LSTM) with dynamic loss weights is proposed. Firstly, a novel multi‐task TCN‐Bi‐LSTM model is designed to extract common features. The separate TCNs, and common Bi‐LSTM layers of the proposed model are designed to extract the temporal features from related wind farms. Secondly, in the pre‐training stage, to optimize the training process of the neural networks, a dynamic loss‐weighting strategy is proposed for multi‐task learning (MTL) to select the most related features, which increase the prediction accuracy by providing a suitable optimization object. Thirdly, the multi‐task TCN‐Bi‐LSTM model is re‐trained based on the samples from the target wind farm. Finally, a dataset of seven wind farms was employed to evaluate the efficiency of the proposed MTL structure and the dynamic loss‐weighting strategy. The result shows that the root mean squared error of the 12‐h short‐term prediction can be decreased by 4.19% compared with the traditional single‐task learning model, which verifies the validity of the proposed multi‐task transfer learning method.

Classification of plant leaf diseases using deep neural networks in color and grayscale images

For a long time, plants have shown a crucial role in our life, manufacturing and industry. However, a large number of diseases have significantly affected to the production of plants. The detection and diagnosis of the diseases are necessary to improve the production of plants. In recent years, the automatic classification of plant diseases using artificial intelligence and computer vision have attracted a huge number of researches. The paper presents a classification method of plant leaf diseases using various Deep Neural Networks (e.g., Alexnet, Resnet-50, Densenet-121). The color features have significantly affected the classification accuracy. Therefore, we analyzed and compared the classification accuracy on two public datasets (Plant village leaf datasets) that consist of color and grayscale images. The method obtained the classification accuracy of 98.08% and 92% on color and grayscale images, respectively. The obtained results showed the effectiveness of the method. Based on the obtained results, the impact of color features to the classification accuracy of various Deep Neural Networks is analyzed in the paper. Moreover, the paper compares the performance of various optimization algorithms during the training process of deep neural networks to classify leaf diseases.

Abstract PO1-23-11: Differential m6A modification identified by direct RNA sequencing in endocrine- resistant and sensitive breast cancer cells

Abstract A major limitation of current adjuvant treatments for the ~ 70% of breast cancer patients with estrogen receptor α–expressing (ER+) primary breast tumors is the development of acquired and adaptive endocrine therapy (ET) resistance and metastatic spread. Chemical modifications of transcribed RNAs alters their cellular location, stability, and, for mRNAs, translation into proteins. N-6 methyladenosine (m6A) is the most common modification of mRNA. We tested the hypothesis that m6A epitranscriptomic changes alters the transcriptome resulting in changes in targets in pathways contributing to the metastatic cascade in ET-resistant breast cancer cells. We used an unbiased approach of direct mRNA sequencing using the Oxford Nanopore MinION platform on polyA-selected mRNA from MCF-7 luminal A breast cancer cells and ER+, ET-resistant LCC9 breast cancer cells treated with vehicle control or 100 nM 4-hydroxytamoxifen (4-OHT) for 24 h. In addition, MCF-7 cells were treated with the METTL3 (the catalytic methyltransferase that adds the methyl group from S-adenosylmethionine (SAM) to the adenosine on the N6 position) inhibitor STM2457 to identify m6A positions and transcripts directly regulated by METTL3 activity. Computational analysis directly identified m6A locations in transcripts. We examined the differential modification rate (DMR) between MCF-7 cells +/STM2457 treatment, between MCF-7 and LCC9 cells with vehicle control treatment or with 4-OHT treatment and within each cell line with 4-OHT treatment. We identified 183 and 3,097 genes with m6A DMRs between vehicle-treated and 4-OHT-treated MCF-7 and LCC9 cells, respectively. For 4-OHT-treated cells, we identified 2,401 and 2,638 genes with m6A DMRs in MCF-7 and LCC9 cells respectively with 1,760 genes in common. We identified many genes with multiple m6A modifications, e.g., AMIGO2 appeared to have 14 m6A sites at DRACH motifs in MCF-7 cells, but none were m6A modified in LCC9 cells. Enrichment by Pathway Maps, Process Networks, and GO Processes were performed. Overall, our data suggest a role for altered m6A positions in transcripts and pathways involved in ET-resistance in breast cancer. Citation Format: Belinda Petri, Bailey Valdes, Kellianne Piell, Eric Rouchka, Carolyn Klinge. Differential m6A modification identified by direct RNA sequencing in endocrine- resistant and sensitive breast cancer cells [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO1-23-11.

A Hybrid Routing Pattern in Human Brain Structural Network Revealed By Evolutionary Computation.

The human brain functional connectivity network (FCN) is constrained and shaped by the communication processes in the structural connectivity network (SCN). The underlying communication mechanism thus becomes a critical issue for understanding the formation and organization of the FCN. A number of communication models supported by different routing strategies have been proposed, with shortest path (SP), random diffusion (DIF), and spatial navigation (NAV) as the most typical, respectively requiring network global knowledge, local knowledge, and both for path seeking. Yet these models all assumed every brain region to use one routing strategy uniformly, ignoring convergent evidence that supports the regional heterogeneity in both terms of biological substrates and functional roles. In this regard, the current study developed a hybrid communication model that allowed each brain region to choose a routing strategy from SP, DIF, and NAV independently. A genetic algorithm was designed to uncover the underlying region-wise hybrid routing strategy (namely HYB). The HYB was found to outperform the three typical routing strategies in predicting FCN and facilitating robust communication. Analyses on HYB further revealed that brain regions in lower-order functional modules inclined to route signals using global knowledge, while those in higher-order functional modules preferred DIF that requires only local knowledge. Compared to regions that used global knowledge for routing, regions using DIF had denser structural connections, participated in more functional modules, but played a less dominant role within modules. Together, our findings further evidenced that hybrid routing underpins efficient SCN communication and locally heterogeneous structure-function coupling.

FAGSO-BP neural network model for settlement prediction of soft soil foundation

To address the problem of difficulty in theoretical formulas and numerical analysis methods to consider the prediction of foundation settlement in the case of multiparameter coupling, the fireworks algorithm (FWA) with a gravity search operator (FAGSO) is introduced into the back propagation (BP) neural network model, and the weight and threshold processes in the neural network are improved by the FAGSO algorithm. A soft soil foundation settlement prediction model based on a FWA optimized BP neural network (FAGSO-BP) with a gravity search operator is proposed. Based on the second phase of the soft foundation treatment project of the Shatian Port Area, Humen City, and Dongguan City, the parameters of pile load height, pore water pressure, and groundwater level were collected and tested. The results indicate that the average relative error of the predicted cumulative settlement value of the soft soil foundation was 0.06%, root mean square error was 7.68, mean absolute error was 4.30, mean absolute percentage error was 1.40%, and mean square error was 59.98. Compared with the BP neural network, GA-BP neural network, and FWA-BP neural network, the error was smaller and the model stability was higher.

Physics-Constraint Variational Neural Network for Wear State Assessment of External Gear Pump.

Most current data-driven prognosis approaches suffer from their uncontrollable and unexplainable properties. To address this issue, this article proposes a physics-constraint variational neural network (PCVNN) for wear state assessment of the external gear pump. First, a response model of the pressure pulsation of the gear pump is constructed via a spectral method, and a compound neural network is utilized to extract features from the pressure pulsation signal. Then, the response model is formulated into an objective function to softly constrain the learning process of the neural network, forcing the learned features to have explicit physics meaning. Meanwhile, to characterize the system uncertainty, the variational inference is utilized to extend a Kullback-Leibler (KL) divergence into the objective function. Finally, the wear state is evaluated based on the distance of learned physics features. Experimental results on an external gear pump validate the merits of the proposed method in explainable representation learning and system uncertainty estimation. It also offers a controllable and explainable perspective to understand the dynamic behavior of the system.