Tracking the information about your manuscript
Communicate with the editorial office
Query manuscript payment status Peer ReviewCollecting, editing, reviewing and other affairs offices
Managing manuscripts
Managing author information and external review Expert Information Office WorkOnline Review
Online Communication with the Editorial Department
Download
Page Views
A Survey on the Construction and Application of Knowledge Graphs Driven by Large Language Models
ZHANG Kunli;WANG Ying;FU Wenhui;ZHU Yongqi;ZHANG Yanli;ZAN Hongying;With the support of large language models, knowledge graphs, characterized by their structured and semantically rich features, enhanced data association and interpretability. This provided new research directions and application potential for complex knowledge reasoning and intelligent decision-making. Therefore, from the perspective of knowledge graphs, the latest progress in the construction and application of knowledge graphs driven by large language models was summarized. Firstly, the new methods for building knowledge graphs were explored from the perspectives of knowledge modeling, information extraction, knowledge integration, and knowledge graph completion. Secondly, the applications of knowledge graphs in augmenting large language models, improving retrieval capabilities, and achieving mutual enhancement with large language models were investigated. Finally, the future directions of combining large language models with knowledge graphs were discussed.
A Prototype Optimization and Refinement Segmentation Network for Few-shot Medical Image Segmentation
WEI Mingjun;HE Haipeng;CHEN Weibin;LIU Yazhi;LI Hui;In order to solve the problem of few-shot medical image segmentation with the distribution shift and local edge details between the support set and the query set, a prototype optimization and refinement segmentation network(PORSNet) for few-shot medical image segmentation was proposed. The network contained a prototype loop iteration module, which suppressed the distribution shift between the initial prototype and the query set and enhanced the expressiveness of the prototype by iteratively performing steps such as initial prototype correction, prototype global perception, and prototype distillation. In addition, a prototype refinement segmentation module was included to further process edge detail information through mask-guided aggregation and feature normalization refinement. Extensive experiments on two widely used few-shot medical public datasets, ABD-MRI and ABD-CT, showed that the proposed PORSNet could use a small number of samples to ensure the segmentation effect with good generalization ability.
Bus Passenger Flow Prediction Based on Multiple Information Attention and Adversarial Graph Convolution
YAN Jianqiang;ZHAO Renqi;GAO Yuan;QU Boting;Aiming at the difficulty of utilizing spatiotemporal dependence relationship in bus passenger flow prediction effectively, a prediction model of passenger flow based on multiple information attention and dynamic adaptive adversarial graph convolutional network was proposed. Firstly, the time feature encoder was used to capture the similarity between passenger flows at different time periods, and point of interest(POI) information of bus stations was incorporated to enhance node feature extraction. Secondly, the dynamic modeling of spatiotemporal dependence was adopted to complete the modeling of non-Euclidean relationships, and the SimAM attention module was utilized to capture the overall differences in passenger flow data at different stations. The experimental results on real bus passenger flow data showed that compared with the best baseline model, the proposed model reduced the average MAE and RMSE of the next 12 time steps by 0.34 and 0.33, respectively, demonstrating its effectiveness and superiority in predicting bus passenger flow.
Domain Adaptation Based on Prompt Learning and Classification Certainty Maximization
DING Meirong;ZHUO Jinxin;LIU Qinglong;LANG Jicong;Domain adaptation faced the issue of complex and variable real-world scenarios, and existing methods mostly focused on optimizing classification consistency while neglecting classification certainty. To address these issues, a network model combining constrastive language-image pre-training(CLIP) with classification certainty maximization was proposed. CLIP, as a multimodal pre-trained model, was pre-trained on a large scale of image-text pairs and possessed strong cross-domain generalization capabilities. By leveraging prompt learning and contrastive learning, the knowledge of the CLIP model was acquired, enabling the model to adapt more complex real-world scenarios. Through the method of classification certainty maximization, a dual-classifier was employed to assess classification consistency and reduce confusion during the model′s inference process. Experiments were conducted on three domain adaptation benchmark datasets: Office-31, Office-Home, and MiniDomainNet. The experimental results indicated that compared with existing advanced methods, the proposed model showed improvements in image classification accuracy across all three datasets.
A Reinforcement Learning Based Approach to Multi-objective Microservice Deployment
ZHANG Menyao;ZHANG Yingxi;ZHENG Wenqi;FENG Guangsheng;In edge computing, microservice architecture could improve data processing efficiency and application response speed, which was suitable for various application scenarios with fast response and frequent interactions. However, existing studies neglected the impact of different interaction frequencies between microservices on the communication overhead. To address this problem, a multi-objective microservice optimal deployment method based on reinforcement learning to improve the performance of microservices in edge environments was proposed. A dual optimization objective model that considers reducing the communication overhead of microservice interactions and balancing the resources of edge nodes was established. Then a deep Q-learning algorithm based on an improved reward mechanism was designed. In order to adapt to the characteristics of shared resources in the process of microservice deployment, a shared reward mechanism was introduced so that the algorithm had better convergence. The experimental results showed that the proposed algorithm could balance the microservice interaction perception and node resource utilization better, and had shorter response time compared with the existing DIM method and Kubernetes default deployment method.
Electric Vehicle Charging and Discharging Strategy Based on Fuzzy Inference and Zoned Dynamic Electricity Price
WANG Qiang;WU Qianfeng;HUANG Bo;WEI Huakun;To address the load fluctuation issues caused by large-scale electric vehicles integration into the power grid, a scheduling strategy for electric vehicle charging and discharging was proposed based on user behavior fuzzy inference and zoned dynamic electricity price. Firstly, a spatiotemporal distribution model of the electric vehicle charging load was constructed based on the travel chain theory, and zoned dynamic electricity price and incentive electricity price models were established according to the node voltage levels. Secondly, users were categorized into rigid users and flexible users based on the real-time state of charge and the required electricity for the next trip. Immediate charging strategies were adopted for rigid users. For flexible users, fuzzy inference theory was utilized to quantify their willingness to respond to charging and discharging. Finally, a multi-objective optimization model was developed with the goal of minimizing the total cost of user charging and discharging and the load variance of distribution network, and maximizing the comprehensive satisfaction of users. And the NSGA-III algorithm was employed to solve the optimization model. The simulation results demonstrated that the proposed strategy could effectively reduce the load variance and peak-valley difference of the power grid, decrease the total cost for users, and significantly improve the voltage quality of distribution network nodes.
[Downloads: 115 ] [Citations: 0 ] [Reads: 0 ] HTML PDF Cite this article
Model Inversion Attack Driven by Frequency Band Decoupling and Dynamic Focus Loss
YANG Jiashuai;WEN Bin;ZHAO Jiateng;ZHOU Shang;To address the problems existing in the model inversion attack based on generative adversarial network, such as frequency band feature coupling, insufficient utilization of label information, and difficulty in optimizing hard samples by loss functions, an enhanced attack method driven by frequency band conditional generator and dynamic focus loss function was proposed. Firstly, the frequency-domain features of the explicitly decoupled input image was achieved through learnable low-pass and high-pass filtering, and multi-scale fusion was performed with the label embedding features to achieve intra-class frequency-domain feature decoupling and precise feature recombination. Secondly, inspired by focus loss,the maximization boundary loss function was improved, and dynamic focal margin loss function was proposed to optimize the search process in the target class space. Experimental results on the CelebA, FFHQ, and FaceScrub face datasets showed that the proposed method significantly improved the performance of model inversion attacks compared to the existing attack methods.
[Downloads: 28 ] [Citations: 0 ] [Reads: 0 ] HTML PDF Cite this article
UAV Small Target Detection Algorithm Based on Improved YOLOv11n
YAO Yue;YANG Xiaohong;GAO Zhiwei;TANG Qiang;To address the challenges including target scale variation,dense occlusion,and background interference in small target detection from UAV perspectives,an detection algorithm based on improved YOLOv11n was proposed.Firstly,a multi-modal feature enhancement attention module was introduced to strengthenthe model′s perception of key regions through integrating multi-dimensional feature inform ation,thereby improving the completeness and discriminability of feature representation.Secondly,an MS＿C3 K2 module was constructed,which utilized convolutional kernels with different receptive fields to pr ocessinput features in parallel,effectively enhancing the ability to extract and fuse both local and globalinform ation.Finally,to further enhance the adaptabilityand dynamism of the fea ture fusion stage,an adaptivegated bi-directional feature pyra mid network was proposed to dynamicallyfuse multi-scale features,enhanc ing the network′s attentionand response capability to key information.Experimental results dem onstratedthat the mAP50 of the improved model reached 33.1%,mAP50-95 reached 18.9%,which was3.4 and 2.3 percentage points higher than th e baseline model,respectively,validating its excellent detectionperformance and robustness in complexUAV sc enarios.
[Downloads: 507 ] [Citations: 0 ] [Reads: 0 ] HTML PDF Cite this article
Event-triggered Control for Time-delay Systems Based on Bumpless Transfer
FU Lei;SUN Yuexing;CUI Shang;Aiming at the event-triggered control problem of bumpless transfer for switched time-delay systems, an event-triggered bumpless transfer strategy and collaborative design method with time-delay dependence were proposed. Firstly, according to the input signal jump suppression requirement, the eventtriggered bumpless transfer performance was defined, and an event-triggered mechanism related to this performance was designed to reduce the control disturbance caused by switching and triggering. Secondly, a multiple Lyapunov function with time-delay terms was constructed, and state-dependent switching laws, event-triggered mechanisms and controllers were co-designed to derive sufficient conditions for the system′s asymptotic stability and H_∞positive lower bound of the event-triggering interval, and the effectiveness of the proposed method compared with the traditional strategy was verified through simulation of an aero-engine model.
[Downloads: 53 ] [Citations: 0 ] [Reads: 0 ] HTML PDF Cite this article
Table Structure Recognition with Dynamic Spatial Perception and Feature Enhancement
LI Zongmin;MA Jinyue;BAI Yun;WANG Xingyu;LIU Yujie;LIU Peigang;Table structure recognition remains highly challenging due to the diversity and complexity of real-world table formats. Existing methods often fail to generalize to different scales and layouts, as they fail to adequately capture layout-specific spatial dependencies and neglect the inherent structural regularities of tables. Consequently, these shortcomings introduce geometric perception bias during feature extraction, thereby limiting the accuracy of cell-relationship modeling and structural prediction. To address these issues, A novel framework was proposed, dynamic spatial perception and feature enhancement(DSPFE), for table structure recognition. A dynamic spatial perception module was introduced to adaptively adjusts the receptive field according to the scale and layout characteristics of each input table,thereby improving spatial adaptability. Furthermore, an aspect-ratio-guided dynamic selection mechanism, coupled with a direction-sensitive feature enhancement module, dynamically selects appropriate enhancement strategies based on directional layout patterns. Residual connections were employed to preserve original structural information while enabling geometric self-adaptation within the feature space. Extensive experiments on benchmark datasets, including PubTabNet and FinTabNet, demonstrate that DSPFE consistently outperforms state-of-the-art methods. Notably, it achieves superior recognition accuracy while maintaining a lightweight architecture, making it well-suited for practical deployment.
[Downloads: 54 ] [Citations: 0 ] [Reads: 0 ] HTML PDF Cite this article
Second Development for Fore Treatment of ABAQUS Using Python Language
ZHONG Tong-sheng~1,WEI Feng~1,WANG Zhi~2,ZHI You-hai~1 (1.Department of Mechanical Engineering,Northwestern Polytechnical University, Xi'an 710072,China;2.Department of Civill Engineering and Architecture, Northwestern Polytechnical University,Xi'an 710072,China)The fore treatment modular of ABAQUS software is developed using Python language.The function and calling procedure of the scenarios ports and target models in the second treatment of ABAQUS are discussed.The model building and ploting grid course of ABAQUS are controlled by the Python scenarios program.Therefore,it improves the fore treatment efficiency and solves the fussy problem that apprears during the manual model building of the model which includes a lot of repeated members.
A New Extreme Learning Machine Optimized by PSO
WANG Jie,BI Hao-yang(School of Electrical Engineering,Zhengzhou University,Zhengzhou 450001,China)Extreme learning machine(ELM) was a new type of feedforward neural network.Compared with traditional single hidden layer feedforward neural networks,ELM possessed higher training speed and smaller error.Due to random input weights and hidden biases,ELM might need numerous hidden neurons to achieve a reasonable accuracy.A new ELM learning algorithm,which was optimized by the particle swarm optimization(PSO),was proposed.PSO algorithm was used to select the input weights and bias of hidden layer,then the output weights could be calculated.To test the validity of proposed method,two simulation experiments were drawn on the approximation curves of the Sinc function.Experimental results showed that the proposed algorithm achieved better performance with less hidden neurons than other similar methods.
A Review on Image Mosaicing Techniques
PEI Hongxing;LIU Jinda;GE Jialong;ZHANG Bin;School of Physics and Engineering,Zhengzhou University;Image mosaicing was analyzed and summarized comprehensively. Based on the development of this technology the background and applications were described generally. The definition and steps of image mosaicing were indicated initially,and the algorithms of image registration were examined by classifying them into several groups. For each algorithm, especially the scale-invariant feature transform( SIFT),the basic theory and characteristics were demonstrated in detail. In the last part,the classifications of image fusion were illustrated and methods of this stage,specially the weighted average,were explained briefly. In addition,the challenges and tendencies of image mosaicing were also pointed out.
Study on the Stability of Aqueous Ferrate(VI) Solution
Jia Handong Ma Ning Sun Hongbin Liu Tongliang Yang Yong (Department of Chemistry, Zhengzhou University, Zhenzhou 450052)The catalytic behaviour of Fe(OH) 3 for the hydrolysis reaction of ferrate(VI) is reported.The effect of temperature, concentration and alkalinity on stability of aqueous ferrate(VI) solutions was studied. It has been descovered that the stability of FeO 2- 4 decreased with increasing amount of Fe(Ⅲ) and with increasing concentration of FeO 2- 4, that the solubility of FeO 2- 4 was not affected by alkalinity at the concentration of OH - below 1.0 mol/L, and that dilute solution of FeO 2- 4 is stable at the normal temperature . These important laws have been used to the synthesis of potassium ferrate(VI) with satisfactory results.
Review on Infrared Dim and Small Target Detection Technology
REN Xiangyang;WANG Jie;MA Tianlei;ZHU Xiaodong;BAI Ke;WANG Jiaqi;With the continuous development of infrared detection technology, the requirements for detection distance were increasingly high. The infrared dim and small target detection technology became the focus of infrared detection. The background and significance of infrared dim and small target detection were reviewed at first. Then the current research status and latest progress of various typical methods in the field of infrared dim and small target detection were trained. After that, experimental comparisons of several different types of infrared dim and small target detection methods were presented. Finally, the research on infrared dim and small target detection technology was summarized and prospected.
Second Development for Fore Treatment of ABAQUS Using Python Language
ZHONG Tong-sheng~1,WEI Feng~1,WANG Zhi~2,ZHI You-hai~1 (1.Department of Mechanical Engineering,Northwestern Polytechnical University, Xi'an 710072,China;2.Department of Civill Engineering and Architecture, Northwestern Polytechnical University,Xi'an 710072,China)The fore treatment modular of ABAQUS software is developed using Python language.The function and calling procedure of the scenarios ports and target models in the second treatment of ABAQUS are discussed.The model building and ploting grid course of ABAQUS are controlled by the Python scenarios program.Therefore,it improves the fore treatment efficiency and solves the fussy problem that apprears during the manual model building of the model which includes a lot of repeated members.
A Review on Image Mosaicing Techniques
PEI Hongxing;LIU Jinda;GE Jialong;ZHANG Bin;School of Physics and Engineering,Zhengzhou University;Image mosaicing was analyzed and summarized comprehensively. Based on the development of this technology the background and applications were described generally. The definition and steps of image mosaicing were indicated initially,and the algorithms of image registration were examined by classifying them into several groups. For each algorithm, especially the scale-invariant feature transform( SIFT),the basic theory and characteristics were demonstrated in detail. In the last part,the classifications of image fusion were illustrated and methods of this stage,specially the weighted average,were explained briefly. In addition,the challenges and tendencies of image mosaicing were also pointed out.
Research on Trusted Cloud Computing Technologies
ZHANG Liqiang;LYU Jianrong;YAN Fei;XIONG Yunfei;With the advantages of high performance, servitization, elastic scale and environmental-friendliness, as a new IT infrastructure, cloud computing has been widely used. Because of its feature of resource outsourcing and resource renting, security and privacy requirements were of great importance. Traditional security technologies were unable to meet the requirements of security in cloud computing. So in recent years, vast security improvements and innovations were proposed in academia and industry. These schemes were used to solve various security problems in cloud computing bottom-up, and built a trusted cloud system architecture in order to achieve a secure and reliable cloud computing. Based on the security threats to cloud computing, the implementations and key technologies of trusted cloud computing were discussed. The advantages and drawbacks of the related works were summarized, and the developing directions of trusted cloud computing were discussed.
Synthesis and Characterization of Mesoporous Molecular Sieves SBA-15
ZHENG Xiu-cheng,YUAN Cheng-yuan,ZHAO Wen-ping,YANG Chun-yan,YE Wen-hao,WANG Xiang-yu(Department of Chemistry,Zhengzhou University,Zhengzhou 450001,China)Mesoporous molecular sieves SBA-15 are synthesized by conventional hydrothermal process using tetraethylorthosilicate(TEOS) as silicon source and triblock copolymer P123 as template.The synthesized samples are characterized by N2 adsorption-desorption,X-ray diffraction (XRD) and transmission electron microscopy(TEM).The results show that with SiO2/P123 ratio increasing,the pore volume and pore size of SBA-15 decrease.With the increasing of reaction solution acidity,the surface area does not change observably but the pore volume and pore size increase.The addition of co-solvents can increase the pore volume and pore size to a certain extent and PhMe3 is the best one.
A Survey on Medical Information Privacy Protection Based on Blockchain
LIU Wei;PENG Yufei;TIAN Zhao;SHENG Zhaoyang;LI Yang;SHE Wei;With the development of medical information, the privacy of medical data has attracted widespread concern among researchers in the process of sharing and accessing. As a decentralized, anonymous, non-tamperable distributed ledger technology, blockchain provided new ideas for solving privacy protection problems in medical scenarios. Firstly, the privacy protection requirements of medical data were listed, and the overall architecture of the blockchain was introduced. Then, the medical information privacy protection technology was introduced in detail, which was divided into data-oriented privacy protection and user-oriented privacy protection. Data-oriented privacy protection was referred to as the protection of sensitive information itself. Encryption-based privacy protection methods, distortion-based privacy protection methods and privacy protection methods based on restricted release were used. User-oriented privacy protection was the privacy protection of data users. It included privacy protection based on access control and transaction anonymity. Finally, the characteristics of various methods were compared and the research status of blockchain in the field of privacy protection was summarized. We prospected the development direction of blockchain in the field of medical information privacy protection was discussed.
Links
Zhengzhou University
Journal of Zhengzhou University
Journal of Tsinghua University(Science and Technology)
Journal of Zhejiang University(Science Edition)
Journal of Jilin University(Science Edition)
Journal of Shandong University(Natural Science)
Journal of Wuhan University(Natural Science Edition)
more..