Loading...

Table of Content

    10 January 2022, Volume 42 Issue 1
    Artificial intelligence
    Unsupervised attributed graph embedding model based on node similarity
    Yang LI, Anbiao WU, Ye YUAN, Linlin ZHAO, Guoren WANG
    2022, 42(1):  1-8.  DOI: 10.11772/j.issn.1001-9081.2021071221
    Asbtract ( )   HTML ( )   PDF (864KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Attributed graph embedding aims to represent the nodes in an attributed graph into low-dimensional vectors while preserving the topology information and attribute information of the nodes. There are lots of works related to attributed graph embedding. However, most of algorithms proposed in them are supervised or semi-supervised. In practical applications, the number of nodes that need to be labeled is large, which makes these algorithms difficult and consume huge manpower and material resources. Above problems were reanalyzed from an unsupervised perspective, and an unsupervised attributed graph embedding algorithm was proposed. Firstly, the topology information and attribute information of the nodes were calculated respectively by using the existing non-attributed graph embedding algorithm and attributes of the attributed graph. Then, the embedding vector of the nodes was obtained by using Graph Convolutional Network (GCN), and the difference between the embedding vector and the topology information and the difference between the embedding vector and attribute information were minimized. Finally, similar embeddings was obtained by the paired nodes with similar topological information and attribute information. Compared with Graph Auto-Encoder (GAE) method, the proposed method has the node classification accuracy improved by 1.2 percentage points and 2.4 percentage points on Cora and Citeseer datasets respectively. Experimental results show that the proposed method can effectively improve the quality of the generated embedding.

    Adaptive deep graph convolution using initial residual and decoupling operations
    Jijie ZHANG, Yan YANG, Yong LIU
    2022, 42(1):  9-15.  DOI: 10.11772/j.issn.1001-9081.2021071289
    Asbtract ( )   HTML ( )   PDF (648KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The traditional Graph Convolutional Network (GCN) and many of its variants achieve the best effect in the shallow layers, and do not make full use of higher-order neighbor information of nodes in the graph. The subsequent deep graph convolution models can solve the above problem, but inevitably generate the problem of over-smoothing, which makes the models impossible to effectively distinguish different types of nodes in the graph. To address this problem, an adaptive deep graph convolution model using initial residual and decoupling operations, named ID-AGCN (model using Initial residual and Decoupled Adaptive Graph Convolutional Network), was proposed. Firstly, the node’s representation transformation as well as feature propagation was decoupled. Then, the initial residual was added to the node’s feature propagation process. Finally, the node representations obtained from different propagation layers were combined adaptively, appropriate local and global information was selected for each node to obtain node representations containing rich information, and a small number of labeled nodes were used for supervised training to generate the final node representations. Experimental result on three datasets Cora, CiteSeer and PubMed indicate that the classification accuracy of ID-AGCN is improved by about 3.4 percentage points, 2.3 percentage points and 1.9 percentage points respectively, compared with GCN. The proposed model has superiority in alleviating over-smoothing.

    Four-layer multiple kernel learning method based on random feature mapping
    Yue YANG, Shitong WANG
    2022, 42(1):  16-25.  DOI: 10.11772/j.issn.1001-9081.2021010171
    Asbtract ( )   HTML ( )   PDF (665KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Since there is no perfect theoretical basis for the selection of kernel function in single kernel network models, and the network node size of Four-layer Neural Network based on Randomly Feature Mapping (FRFMNN) is excessively large, a Four-layer Multiple Kernel Neural Network based on Randomly Feature Mapping (MK-FRFMNN) algorithm was proposed. Firstly, the original input features were transformed into randomly mapped features by a specific random mapping algorithm. Then, multiple basic kernel matrices were generated through different random kernel mappings. Finally, the synthetic kernel matrix formed by basic kernel matrices was linked to the output layer through the output weights. Since the weights of random mapping of original features were randomly generated according to the random continuous sampling probability distribution randomly, without the need of updates of the weights, and the weights of the output layer were quickly solved by the ridge regression pseudo inverse algorithm, thus avoiding the time-consuming training process of the repeated iterations. Different random weight matrices were introduced into the basic kernel mapping of MK-FRFMNN. the generated synthetic kernel matrix was able to not only synthesize the advantages of various kernel functions, but also integrate the characteristics of various random distribution functions, to obtain better feature selection and expression effect in the new feature space. Theoretical and experimental analyses show that, compared with the single kernel models such as Broad Learning System (BLS) and FRMFNN, MK-FRMFNN model has the node size reduced by about 2/3 with stable classification performance; compared with mainstream multiple kernel models, MK-FRMFNN model can learn large sample datasets, and has better performance in classification.

    Opinion leader recognition algorithm based on K-core decomposition in social networks
    Meizi LI, Yifei MI, Qian ZHANG, Bo ZHANG
    2022, 42(1):  26-35.  DOI: 10.11772/j.issn.1001-9081.2021010138
    Asbtract ( )   HTML ( )   PDF (1476KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    In view of the high computational complexity of opinion leader mining in social networks, an opinion leader recognition algorithm based on K-core decomposition, named CandidateRank (CR), was proposed. Firstly, the opinion leader candidate set in a social network was obtained based on K-core decomposition method, so as to reduce the data size of opinion leader recognition. Then, a user similarity concept including location similarity and neighbor similarity was proposed, and the user similarity was calculated by K-core value, the number of entries, average K-core change rate and the number of user followers, and the global influence of the user in the candidate set was calculated according to the user similarity. Finally, opinion leaders were recognized by ranking users in the opinion leader candidate set by the global influence. In the experiment, two evaluation indexes of user influence predicted by Independent Cascade Model (ICM) and centrality were used to evaluate the opinion leader set selected by the proposed algorithm on three real datasets with different sizes. The results show that the proposed algorithm has the average user influence for the selected Top-15 users of 21.442, which is higher than those of the other three algorithms. In addition, compared to four K-core-related algorithms in correlation index, the results show that CandidateRank algorithm performs better in general. In summary, CandidateRank algorithm improves the accuracy while reducing the computational complexity.

    Sparrow search algorithm based on Sobol sequence and crisscross strategy
    Yuxian DUAN, Changyun LIU
    2022, 42(1):  36-43.  DOI: 10.11772/j.issn.1001-9081.2021010187
    Asbtract ( )   HTML ( )   PDF (771KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    For the shortcomings of falling into the local optimum easily and slow convergence in Sparrow Search Algorithm (SSA), a Sparrow Search Algorithm based on Sobol sequence and Crisscross strategy (SSASC) was proposed. Firstly, the Sobol sequence was introduced in the initialization stage to enhance the diversity and ergodicity of the population. Secondly, the nonlinear inertia weight in exponential form was proposed to improve the convergence efficiency of the algorithm. Finally, the crisscross strategy was applied to improve the algorithm. In specific, the horizontal crossover was used to enhance the global search ability, while the vertical crossover was used to maintain the diversity of the population and avoid the algorithm from trapping into the local optimum. Thirteen benchmark functions were selected for simulation experiments, and the performance of the algorithm was evaluated by Wilcoxon rank sum test and Friedman test. In comparison experiments with other metaheuristic algorithms, the mean and standard deviation generated by SSASC are always better than other algorithms when the benchmark functions extending from 10 dimensions to 100 dimensions. Experimental results show that SSASC achieves certain superiority in both convergence speed and solution accuracy.

    Constrained differentiable neural architecture search in optimized search space
    Jianming LI, Bin CHEN, Zhiwei JIANG, Jian QIN
    2022, 42(1):  44-49.  DOI: 10.11772/j.issn.1001-9081.2021010170
    Asbtract ( )   HTML ( )   PDF (603KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Differentiable ARchiTecture Search (DARTS) can design neural network architectures efficiently and automatically. However, there is a performance “wide gap” between the construction method of super network and the design of derivation strategy in it. To solve the above problem, a differentiable neural architecture search algorithm with constraint in optimal search space was proposed. Firstly, the training process of the super network was analyzed by using the architecture parameters associated with the candidate operations as the quantitative indicators, and it was found that the invalid candidate operation none occupied the architecture parameter with the maximum weight in deviation architecture, which caused that architectures obtained by the algorithm had poor performance. Aiming at this problem, an optimized search space was proposed. Then, the difference between the super network of DARTS and derivation architecture was analyzed, the architecture entropy was defined based on architecture parameters, and this architecture entropy was used as the constraint of the objective function of DARTS, so as to promote the super network to narrow the difference with the derivation strategy. Finally, experiments were conducted on CIFAR-10 dataset. The experimental results show that the searched architecture by the proposed algorithm achieved 97.17% classification accuracy in these experiments, better than the comparison algorithms in accuracy, parameter quantity and search time comprehensively. The proposed algorithm is effective and improves classification accuracy of searched architecture on CIFAR-10 dataset.

    Academic journal contribution recommendation algorithm based on author preferences
    Yongfeng DONG, Xiangqian QU, Linhao LI, Yao DONG
    2022, 42(1):  50-56.  DOI: 10.11772/j.issn.1001-9081.2021010185
    Asbtract ( )   HTML ( )   PDF (605KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    In order to solve the problem that the algorithms of publication venue recommendation always consider the text topics or the author’s history of publications separately, which leads to the low accuracy of publication venue recommendation results, a contribution recommendation algorithm of academic journal based on author preferences was proposed. In this algorithm, not only the text topics and the author’s history of publications were used together, but also the potential relationship between the academic focuses of publication venues and time were explored. Firstly, the Latent Dirichlet Allocation (LDA) topic model was used to extract the topic information of the paper title. Then, the topic-journal and time-journal model diagrams were established, and the Large-scale Information Network Embedding (LINE) model was used to learn the embedding of graph nodes. Finally, the author’s subject preferences and history of publication records were fused to calculate the journal composite scores, and the publication venue recommendation for author to contribute was realized. Experimental results on two public datasets, DBLP and PubMed, show that the proposed algorithm has better recall under different list lengths of recommended publication venues compared to six algorithms such as Singular Value Decomposition (SVD), DeepWalk and Non-negative Matrix Factorization (NMF). The proposed algorithm maintains high accuracy while requiring less information from papers and knowledge bases, and can effectively improve the robustness of publication venue recommendation algorithm.

    Text multi-label classification method incorporating BERT and label semantic attention
    Xueqiang LYU, Chen PENG, Le ZHANG, Zhi’an DONG, Xindong YOU
    2022, 42(1):  57-63.  DOI: 10.11772/j.issn.1001-9081.2021020366
    Asbtract ( )   HTML ( )   PDF (577KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Multi-Label Text Classification (MLTC) is one of the important subtasks in the field of Natural Language Processing (NLP). In order to solve the problem of complex correlation between multiple labels, an MLTC method TLA-BERT was proposed by incorporating Bidirectional Encoder Representations from Transformers (BERT) and label semantic attention. Firstly, the contextual vector representation of the input text was learned by fine-tuning the self-coding pre-training model. Secondly, the labels were encoded individually by using Long Short-Term Memory (LSTM) neural network. Finally, the contribution of text to each label was explicitly highlighted with the use of an attention mechanism in order to predict the multi-label sequences. Experimental results show that compared with Sequence Generation Model (SGM) algorithm, the proposed method improves the F value by 2.8 percentage points and 1.5 percentage points on the Arxiv Academic Paper Dataset (AAPD) and Reuters Corpus Volume I (RCV1)-v2 public dataset respectively.

    Three-stage question answering model based on BERT
    Yu PENG, Xiaoyu LI, Shijie HU, Xiaolei LIU, Weizhong QIAN
    2022, 42(1):  64-70.  DOI: 10.11772/j.issn.1001-9081.2021020335
    Asbtract ( )   HTML ( )   PDF (918KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The development of pre-trained language models has greatly promoted the progress of machine reading comprehension tasks. In order to make full use of shallow features of the pre-trained language model and further improve the accuracy of predictive answer of question answering model, a three-stage question answering model based on Bidirectional Encoder Representation from Transformers (BERT) was proposed. Firstly, the three stages of pre-answering, re-answering and answer-adjusting were designed based on BERT. Secondly, the inputs of embedding layer of BERT were treated as shallow features to pre-generate an answer in pre-answering stage. Then, the deep features fully encoded by BERT were used to re-generate another answer in re-answering stage. Finally, the final prediction result was generated by combining the previous two answers in answer-adjusting stage. Experimental results on English dataset Stanford Question Answering Dataset 2.0 (SQuAD2.0) and Chinese dataset Chinese Machine Reading Comprehension 2018 (CMRC2018) of span-extraction question answering task show that the Exact Match (EM) and F1 score (F1) of the proposed model are improved by the average of 1 to 3 percentage points compared with those of the similar baseline models, and the model has the extracted answer fragments more accurate. By combining shallow features of BERT with deep features, this three-stage model extends the abstract representation ability of BERT, and explores the application of shallow features of BERT in question answering models, and has the characteristics of simple structure, accurate prediction, and fast speed of training and inference.

    Customs declaration good classification algorithm based on hierarchical multi-task BERT
    Qiming RUAN, Yi GUO, Nan ZHENG, Yexiang WANG
    2022, 42(1):  71-77.  DOI: 10.11772/j.issn.1001-9081.2021010122
    Asbtract ( )   HTML ( )   PDF (697KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    In the customs good declaration scenarios, a classification model needs to be used to categorize the goods into uniform Harmonized System (HS) codes. However, the existing customs good classification models ignore the location information of words in the text to be classified, while the HS codes are in tens of thousands, which leads to problems such as class vector sparsity and slow convergence of the model.To address the above problems, a classification model based on Hierarchical Multi-task Bidirectional Encoder Representation from Transformers (HM-BERT) was proposed by combining the manual hierarchical classification strategy in real business scenarios and making full use of the hierarchical structure feature of HS codes. In one aspect, the dynamic word vector of Bidirectional Encoder Representation from Transformers (BERT) model was used to obtain the location information in the text of customs declaration goods. In other aspect, the accuracy and convergence of categorization were improved by making full use of the category information of different levels of HS codes to perform multi-task training of BERT model. In the effectiveness verification of the proposed model on the 2019 customs declaration dataset of a domestic customs service provider, HM-BERT model improves 2 percentage points in accuracy with faster training speed compared to BERT model, and improves 7.1 percentage points in accuracy compared with H (Hierarchical)-fastText. Experimental results show that HM-BERT model can effectively improve the classification effect of customs declaration goods.

    Automatic construction method of knowledge forest for electronic case files
    Yincen QU, Yinliang ZHAO, Chongchong JIU, Shuo LIU
    2022, 42(1):  78-86.  DOI: 10.11772/j.issn.1001-9081.2021020267
    Asbtract ( )   HTML ( )   PDF (1017KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The read of various contents of case files suffers from information overload and knowledge disorientation. To solve this problem, an automatic construction method of knowledge forest for electronic case files was proposed with the topic facet trees and the cognitive relationships between topics as the intellectualized representation of the case files. Firstly, different types of files were classified and divided into multiple fragments of single topic by the fragmentation preprocessing of the case files. Then, different information extraction methods were adopted for different fragments, and knowledge fusion was used to merge the synonymous information. After that, the topic faceted trees were constructed by combining the ontology structures and rules and the topic relationships were extracted. Finally, the topic faceted trees and the topic relationships constructed by the knowledge forest were stored in the database to realize the visualization of the knowledge forest. Experimental results show that the proposed method can display the case file information completely and accurately, organize scattered knowledge fragments together with complex case file topics, making it possible to achieve the reading file goal by selecting some case file topics and a small number of case file fragments, and alleviate the burden of complete browsing case file contents to realize the file reading task.

    Encoding-decoding relationship extraction model based on criminal Electra
    Xiaopeng WANG, Yuanyuan SUN, Hongfei LIN
    2022, 42(1):  87-93.  DOI: 10.11772/j.issn.1001-9081.2021020272
    Asbtract ( )   HTML ( )   PDF (723KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the problem that the model in the judicial field relation extraction task does not fully understand the context of sentence and has weak recognition ability of overlapping relations, based on Criminal-Efficiently learning an encoder that classi?es token replacements accurately (CriElectra), an encoding-decoding relationship extraction model was proposed. Firstly, referred to the training method of Chinese Electra, CriElectra was trained on one million criminal dataset. Then, the word vectors of CriElectra were added to Bidirectional Long Short-Term Memory (BiLSTM) model for feature extraction of judicial texts. Finally, the vector clustering was performed to the features through Capsule Network (CapsNet), so that the relationships between entities were extracted. Experimental results show that on the self-built relationship dataset of intentional injury crime, compared with the pre-trained language model based on Chinese Electra, CriElectra has retraining process on judicial texts to make the learned word vectors contain richer domain information, and the F1-score increased by 1.93 percentage points. Compared with the model based on pooling clustering, CapsNet can effectively prevent the loss of spatial information by vector operation and improve the recognition ability of overlapping relationships, which increases the F1-score by 3.53 percentage points.

    Data science and technology
    Survey of high utility pattern mining on dynamic data
    Zhihui SHAN, Meng HAN, Qiang HAN
    2022, 42(1):  94-108.  DOI: 10.11772/j.issn.1001-9081.2021071290
    Asbtract ( )   HTML ( )   PDF (1668KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    High Utility Pattern Mining (HUPM) provides details about items to let users make better economic decisions by considering the numbers of purchase and the unit profits of items. Since most HUPM algorithms are applied in static databases, which are inconsistent with real-world scenarios where data is constantly generated, HUIM algorithms on dynamic data have been proposed in recent years. Firstly, the HUPM algorithms on incremental data, data stream, dynamic deletion data and dynamic modification data as well as the integrated high utility patterns (such as high utility sequential patterns, average high utility patterns, and top-k high utility patterns) mining algorithms were summarized. Secondly, the algorithms that handled different types of data, including dynamic profit data, dynamic sequence data and other data types, were summed up. Thirdly, the HUPM algorithms were classified and summarized from the perspectives of data structure, pruning strategy, window model, advantages and disadvantages. Finally, aiming at the lack in the current research, the research directions of HUPM algorithm on dynamic data in the next step were proposed.

    Dynamic relevance based feature selection algorithm
    Yongbo CHEN, Qiaoqin LI, Yongguo LIU
    2022, 42(1):  109-114.  DOI: 10.11772/j.issn.1001-9081.2021010128
    Asbtract ( )   HTML ( )   PDF (445KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    By removing irrelevant features from the original dataset and selecting good feature subsets, feature selection can avoid the curse of dimensionality and improve the performance of learning algorithm.In the process of feature selection, only the dynamically change information between the selected features and classes is considered, and interaction relevance between the candidate features and the selected features is ignored by Dynamic Change of Selected Feature with the class (DCSF) algorithm. To solve this problem, a Dynamic Relevance based Feature Selection (DRFS) algorithm was proposed. In the proposed algorithm, conditional mutual information was used to measure the conditional relevance between the selected features and classes, and interaction information was used to measure the synergy brought by the candidate features and the selected features, so as to select relevant features and remove redundant features then obtain good feature subsets. Simulation results show that, compared with existing algorithms, the proposed algorithm can effectively improve classification accuracy of feature selection.

    Low-rank representation subspace clustering method based on Hessian regularization and non-negative constraint
    Lili FAN, Guifu LU, Ganyi TANG, Dan YANG
    2022, 42(1):  115-122.  DOI: 10.11772/j.issn.1001-9081.2021071181
    Asbtract ( )   HTML ( )   PDF (661KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Focusing on the issue that the Low-Rank Representation (LRR) subspace clustering algorithm does not consider the local structure of the data and may cause the loss of local similar information during learning, a Low-Rank Representation subspace clustering algorithm based on Hessian regularization and Non-negative constraint (LRR-HN) was proposed to explore the global and local structure of the data. Firstly, the good speculative ability of Hessian regularization was used to maintain the local manifold structure of the data, so that the local topological structure of the data was more expressive. Secondly, considering that the obtained coefficient matrix often has positive and negative values, and the negative values often have no practical significance, non-negative constraints were introduced to ensure the effectiveness of the model solution and make it more meaningful in the description of the local structure of the data. Finally, the low-rank representation of the global structure of the data was sought by minimizing the nuclear norm, so as to cluster high-dimensional data better. In addition, an effective algorithm for solving LRR-HN was designed by using the linearized alternating direction method with adaptive penalty, and the proposed algorithm was evaluated by ACcuracy (AC) and Normalized Mutual Information (NMI) on some real datasets. In the experiments with clusters number 20 on ORL dataset, compared with LRR algorithm, LRR-HN has the AC and NMI increased by 11% and 9.74% respectively, and compared with Adaptive Low-Rank Representation (ALRR) algorithm, LRR-HN has the AC and NMI increased by 5% and 1.05% respectively. Experimental results show that the LRR-HN has great improvement in AC and NMI compared with some existing algorithms, and has the excellent clustering performance.

    Dynamic weighted ensemble classification algorithm based on accuracy climbing
    Xiaojuan LI, Meng HAN, Le WANG, Ni ZHENG, Haodong CHENG
    2022, 42(1):  123-131.  DOI: 10.11772/j.issn.1001-9081.2021071234
    Asbtract ( )   HTML ( )   PDF (992KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    In the traditional ensemble classification algorithm, the ensemble number is generally set to a fixed value, which may lead to a low classification accuracy. Aiming at this problem, an accuracy Climbing Ensemble Classification Algorithm (C-ECA) was proposed. Firstly, the base classifiers was no longer replaced the same number of base classifiers with the worst performance, but updated based on the accuracy in this algorithm, and then the optimal ensemble number was determined. Secondly, on the basis of C-ECA, a Dynamic Weighted Ensemble Classification Algorithm based on Climbing (C-DWECA) was proposed. When the base classifier was trained on the data stream with different features, the best weight of the base classifier was able to be obtained by a weighting function proposed in this algorithm, thereby improving the performance of the ensemble classifier. Finally, in order to detect the concept drift earlier and improve the final accuracy, Fast Hoffding Drift Detection Method (FHDDM) was adopted. Experimental results show that the accuracy of C-DWECA can reach up to 97.44%, and the average accuracy of the proposed algorithm is about 40% higher than that of Adaptable Diversity-based Online Boosting (ADOB) algorithm, and is also better than those of other comparison algorithms such as Leveraging Bagging (LevBag) and Adaptive Random Forest (ARF).

    Parallel pivoted subgraph matching with multiple coding trees on GPU
    Yang WANG, Shijie JIANG, Yucong CAO, Chuanwen LI
    2022, 42(1):  132-139.  DOI: 10.11772/j.issn.1001-9081.2021071219
    Asbtract ( )   HTML ( )   PDF (877KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The subgraph isomorphism problem is a Non-deterministic Polynomial (NP)-complete problem, and the pivoted subgraph isomorphism is a special subgraph isomorphism problem. There are many existing efficient subgraph isomorphism algorithms, but there is no GPU-based search algorithm for the pivoted subgraph isomorphism problem at present, and a large number of unnecessary intermediate results will be generated when the pivoted subgraph matching problem is solved by the existing subgraph isomorphism algorithms. Therefore, a GPU-based pivoted subgraph isomorphism algorithm was proposed. Firstly, through a novel coding tree method, nodes were encoded by the combination of node labels, degrees and the structural features of node neighbors. And the query graph nodes were pruned on GPU in parallel, so that the size of search space tree generated by the data graph candidate nodes was significantly reduced. Then, the candidate nodes of the query graph node were visited level by level, and the unsatisfied nodes were filtered out. Finally, the obtained subgraph was verified whether it was an isomorphic subgraph of the query graph, and the search of pivoted subgraph isomorphism was realized efficiently. Experimental results show that compared with GPU-friendly Subgraph Matching (GpSM) algorithm, the proposed algorithm has the execution time reduced by one-half, and the proposed algorithm can efficiently perform the pivoted subgraph isomorphism search with scalability. The proposed pivoted subgraph isomorphism algorithm can reduce the time required to solve the pivoted subgraph isomorphism problem, while reducing GPU memory consumption and improving the performance of algorithm.

    Design and implementation of multi-version concurrency control on embedded database SQLite
    Ziqi JING, Zhaonian ZOU
    2022, 42(1):  140-147.  DOI: 10.11772/j.issn.1001-9081.2021071217
    Asbtract ( )   HTML ( )   PDF (785KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    In order to solve the problem of low concurrency performance of embedded database SQLite, a concurrency control design based on Multi-Version Concurrency Control (MVCC) was proposed. Firstly, the version division method of SQLite database based on committed write transactions was designed, and the header field of data records was redesigned to divide the visibility of records under different version accesses. Then, based on the original structure of SQLite, the operations such as add, delete, check and change and the index structure were modified to make the database work under MVCC. Finally, a manual recycling mechanism was provided to handle old version data. Experiments were designed to compare and test the performance difference between SQLite-MVCC database obtained by the above design and SQLite database. It can be seen that in the state of high concurrency, SQLite-MVCC database can complete 70% more transactions in the same time. Experimental results verify that the proposed design can effectively improve the concurrency performance of SQLite, to meet the needs in concurrent case.

    Cyber security
    Survey of anonymity and tracking technology in Monero
    Dingkang LIN, Jiaqi YAN, Nandeng BA, Zhenhao FU, Haochen JIANG
    2022, 42(1):  148-156.  DOI: 10.11772/j.issn.1001-9081.2021020296
    Asbtract ( )   HTML ( )   PDF (723KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Virtual digital currency provides a breeding ground for terrorist financing, money laundering, drug trafficking and other criminal activities. As a representative emerging digital currency, Monero has a universally acknowledged high anonymity. Aiming at the problem of using Monroe anonymity to commit crimes, Monero anonymity technology and tracking technology were explored as well as the research progresses were reviewed in recent years, so as to provide technical supports for effectively tackling the crimes based on blockchain technology. In specific, the evolution of Monero anonymity technology was summarized, and the tracking strategies of Monero anonymity technology in academic circles were sorted out. Firstly, in the anonymity technologies, ring signature, guaranteed unlinkability (one-off public key), guaranteed untraceability, and the important version upgrading for improving anonymity were introduced. Then, in tracking technologies, the attacks such as zero mixin attack, output merging attack, guess-newest attack, closed set attack, transaction flooding attack, tracing attacks from remote nodes and Monero ring attack were introduced. Finally, based on the analysis of anonymity technologies and tracking strategies, four conclusions were obtained: the development of anonymity technology and the development of tracking technology of Monero promote each other; the application of Ring Confidential Transactions (RingCT) is a two-edged sword, which makes the passive attack methods based on currency value ineffective, and also makes the active attack methods easier to succeed; output merging attack and zero mixin attack complement each other; Monero’s system security chain still needs to be sorted out.

    Multiparty quantum key agreement protocol based on logical single particle
    Shengwei XU, Jie KANG
    2022, 42(1):  157-161.  DOI: 10.11772/j.issn.1001-9081.2021010176
    Asbtract ( )   HTML ( )   PDF (456KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The influence of collective noise on quantum cryptographic protocols can not be ignored, however there are few Multiparty Quantum Key Agreement (MQKA) protocols that can resist collective noise. In order to resist the influence of collective noise, two sets of logical unitary operators were proposed for the logical single particle which can resist collective-dephasing noise and the logical single particle which can resist collective rotation noise respectively, so that after acting them on the logical single particles, two unitary operators did not change the measurement base, while the other two would change the measurement base. Based on this property, an MQKA protocol was proposed. Firstly, a logical single particle was transmited to the next participant by each participant.Then, the logical single particle was encrypted by all other participants and returned to the participant to form a “ring”. Finally, the shared key was obtained by measurement. Security analysis shows that the proposed protocol can resist intercept-resend attack, entanglement-measurement attack and participant attack. Efficiency analysis shows that this protocol achieves higher qubit efficiency.

    Nonlinear scrambling diffusion synchronization image encryption based on dynamic network
    Yuan GUO, Xuewen WANG, Chong WANG, Jinlin JIANG
    2022, 42(1):  162-170.  DOI: 10.11772/j.issn.1001-9081.2021071220
    Asbtract ( )   HTML ( )   PDF (3822KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The traditional image encryption with scrambling-diffusion structure is usually divided into two independent steps of scrambling and diffusion, which are easy to be cracked separately, and the encryption process has weak nonlinearity, resulting in poor security of the algorithm. Therefore, a scrambling diffusion synchronous image encryption algorithm with strong nonlinearity was proposed. Firstly, a new sine-cos chaotic mapping was constructed to broaden the range of control parameters and improve the randomness of sequence distribution. Then, the exclusive-OR sum of plaintext pixels and chaotic sequence was used as the initial chaotic value to generate chaotic sequence, and this chaotic sequence was used to construct the network structures of different pixels of different plaintexts. At the same time, the diffusion value was used to dynamically update the network value to make the network dynamic. Finally, the single pixel serial scrambling-diffusion was used to generate cross-effect between scrambling and diffusion,and the overall synchronization of scrambling and diffusion, so as to effectively resist separation attacks. In addition, the pixel operations were transferred according to the network structure, which made the serial path nonlinear and unpredictable, thereby ensuring the nonlinearity and security of the algorithm. And the adjacent node pixels sum was used to perform dynamic diffusion in order to improve the correlation of the plaintext. Experimental results show that the proposed algorithm has high encryption security, strong plaintext sensitivity, and is particularly effective in anti-statistical attack, anti-differential attack and anti-plaintext attack.

    Reversible data hiding method based on high-order bit-plane redundancy
    Cong XU, Xingtian WANG, Yongpeng TAO
    2022, 42(1):  171-177.  DOI: 10.11772/j.issn.1001-9081.2021020237
    Asbtract ( )   HTML ( )   PDF (1374KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Focused on the problems of low hiding capacity and poor quality of decrypted labeled images in the existing Reversible Data Hiding in Encrypted Image (RDHEI) methods, a new RDHEI method based on high-order bit-plane redundancy was proposed. Firstly, the original image was encrypted in blocks by Logistic mapping, and the redundancy of the high-order bit-plane of the pixels in the blocks was retained. Secondly, according to the rule of whether the numbers of high-order bits and low-order bits in the block were the same, the encrypted image blocks were divided into embeddable blocks and non-embeddedable blocks, and the low-order bit value of the pixel was replaced with the corresponding high-order bit value in the embeddable blocks, so that the high-order bit-plane redundancy was transferred to the low-order bit-plane. Finally, the confidential information was embedded in the embedding space vacated in the inner-block low-order bit-plane. After that, the operations of data extraction, image decryption and image lossless recovery were realized by the receiver with the key. In the simulation experiments on 6 images in the USC-SIPI standard image library, when the number of high-order bit-planes is equal to 3, the proposed method has the average embedding rate of the image of 1.73 bpp, and the average Peak Signal-to-Noise Ratio (PSNR) of the marked image after direct decryption reaches 47.20 dB. The experimental results show that the proposed method not only increases the information embedding capacity of the encrypted image, but also increases the PSNR value of the labeled image after direct decryption.

    Memory combined feature classification method based on multiple BP neural networks
    Jialiang DUAN, Guoming CAI, Kaiyong XU
    2022, 42(1):  178-182.  DOI: 10.11772/j.issn.1001-9081.2021010199
    Asbtract ( )   HTML ( )   PDF (563KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The memory data will change after occurring the attack behaviors, and benchmark measurement used by the traditional integrity measurement system has the problems of low detection rate and lack of flexibility. Aiming at the above problems, a memory combined feature classification method based on multiple Back Propagation (BP) neural networks was proposed. Firstly, the feature value of the memory data was extracted by Measuring Object Extraction Algorithm (MOEA). Then, the model was trained by different BP neural networks. Finally, a BP neural network was used to collect the obtained data and calculate the safety status score of the operating system. Experimental results show that compared with the traditional integrity measurement system using benchmark measurement, the proposed method has much higher accuracy and universality, and the proposed method has a detection accuracy of 98.25%, which is higher than those of Convolutional Neural Network (CNN), K-Nearest Neighbor (KNN) algorithm and single BP neural network, verifying the proposed method can detect attack behaviors more accurately. The proposed method has the model training time about 1/3 of the traditional single BP neural network, and also has the model training speed improved compared with similar models.

    Medical electronic record sharing scheme based on sharding-based blockchain
    Li LI, Yi WU, Zhikun YANG, Yunpeng CHEN
    2022, 42(1):  183-190.  DOI: 10.11772/j.issn.1001-9081.2021010107
    Asbtract ( )   HTML ( )   PDF (869KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the limited scalability of medical data sharing based on traditional blockchains, a scale-out and sharing scheme of blockchain based on sharding technology was proposed. Firstly, the periodic network sharding was performed based on the jump consistent hash algorithm, and the risk of Sybil attacks in a single shard was greatly reduced by randomly dividing the network nodes. Then, the Scalable decentralized Trust inFrastructure for Blockchains (SBFT) consensus protocol was used in the shards to reduce the high communication complexity of the Pratic Byzantic Fault Torent (PBFT) consensus protocol, and the two-layer architecture was used between the physical multi-chain of shards and the logical single chain of the main chain to reduce the storage pressure of the members of shards. Finally, a multi-keyword association retrieval searchable encryption sharing scheme based on Public key Encryption with Conjunctive field Keyword Search (PECKS) was proposed on the medical consortium blockchain, so as to improve the patients’ control over their sensitive data, and realize the fine-grained search of sensitive data under encryption. Through performance analysis, it can be seen that under the parallel sharding structure, the throughput of blockchain is significantly increased with the increase of shards, and the retrieval efficiency is also significantly improved. Experimental results show that the proposed scheme can greatly improve the efficiency and scalability of the blockchain system.

    Advanced computing
    Average consensus tracking of multi-agent system with time-varying reference inputs
    Yu ZHANG, Chenglin LIU
    2022, 42(1):  191-197.  DOI: 10.11772/j.issn.1001-9081.2021010197
    Asbtract ( )   HTML ( )   PDF (812KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the dynamic average consensus tracking problem of multi-agent systems with time-varying reference inputs, a proportional-integral consensus tracking algorithm was proposed. In the scenario of communication data between multi-agents being quantized, the average consensus tracking problem based on quantization was studied. Firstly, on the basis of the integral algorithm, a proportional link was introduced to make agents to track the average value of the reference inputs better by communicating with neighborhood agents under the constraints of the control agreement. Then, under the fixed, strongly connected and balanced topology structure, sufficient conditions for the multi-agent system asymptotically tracking to the average value of time-varying reference inputs without and with quantized information transmission data were obtained by using matrix analysis and Routh criteria respectively. Finally, numerical simulations verify the accuracy of the results and confirm the effectiveness of the proposed algorithm.

    Data center server energy consumption optimization algorithm combining XGBoost and Multi-GRU
    Mingyao SHEN, Meng HAN, Shiyu DU, Rui SUN, Chunyan ZHANG
    2022, 42(1):  198-208.  DOI: 10.11772/j.issn.1001-9081.2021071291
    Asbtract ( )   HTML ( )   PDF (1169KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    With the rapid development of cloud computing technology, the number of data centers have increased significantly, and the subsequent energy consumption problem gradually become one of the research hotspots. Aiming at the problem of server energy consumption optimization, a data center server energy consumption optimization combining eXtreme Gradient Boosting (XGBoost) and Multi-Gated Recurrent Unit (Multi-GRU) (ECOXG) algorithm was proposed. Firstly, the data such as resource occupation information and energy consumption of each component of the servers were collected by the Linux terminal monitoring commands and power consumption meters, and the data were preprocessed to obtain the resource utilization rates. Secondly, the resource utilization rates were constructed in series into a time series in vector form, which was used to train the Multi-GRU load prediction model, and the simulated frequency reduction was performed to the servers according to the prediction results to obtain the load data after frequency reduction. Thirdly, the resource utilization rates of the servers were combined with the energy consumption data at the same time to train the XGBoost energy consumption prediction model. Finally, the load data after frequency reduction were input into the trained XGBoost model, and the energy consumption of the servers after frequency reduction was predicted. Experiments on the actual resource utilization data of 6 physical servers showed that ECOXG algorithm had a Root Mean Square Error (RMSE) reduced by 50.9%, 31.0%, 32.7%, 22.9% compared with Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) network, CNN-GRU and CNN-LSTM models, respectively. Meanwhile, compared with LSTM, CNN-GRU and CNN-LSTM models, ECOXG algorithm saved 43.2%, 47.1%, 59.9% training time, respectively. Experimental results show that ECOXG algorithm can provide a theoretical basis for the prediction and optimization of server energy consumption optimization, and it is significantly better than the comparison algorithms in accuracy and operating efficiency. In addition, the power consumption of the server after the simulated frequency reduction is significantly lower than the real power consumption, and the effect of reducing energy consumption is outstanding when the utilization rates of the servers are low.

    Q-table initialization approach for safe exploration based on factorization machine
    Bosen ZENG, Yong ZHONG, Xianhua NIU
    2022, 42(1):  209-214.  DOI: 10.11772/j.issn.1001-9081.2021020239
    Asbtract ( )   HTML ( )   PDF (873KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    In order to solve the problem that most exploration/exploitation strategies of reinforcement learning ignore the risk brought by the agent action selection with random components in exploration process, a Q-table initialization approach based on Factorization Machine (FM) was proposed for safe exploration. Firstly, the explored Q-values were introduced as prior knowledge, and then FM was used to build the model of potential interaction between states and actions in the prior knowledge. Finally, the unknown Q-values in Q-table were predicted based on this model to further guide the exploration of the agents. A/B testing was conducted in the grid reinforcement learning environment Cliffwalk of OpenAI Gym. The number of bad exploration episodes of Boltzmann and Upper Confidence Bound (UCB) exploration/exploitation strategies based on the proposed approach are reduced by 68.12% and 89.98% respectively. Experimental results show that the proposed approach improves the safety of exploration, and accelerates the convergence at the same time.

    Discrete manta ray foraging optimization algorithm and its application in spectrum allocation
    Dawei WANG, Xinhao LIU, Zhu LI, Bin LU, Aixin GUO, Guoqiang CHAI
    2022, 42(1):  215-222.  DOI: 10.11772/j.issn.1001-9081.2021020238
    Asbtract ( )   HTML ( )   PDF (671KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the problem of spectrum allocation based on maximizing network benefit in cognitive radio and the fact that Manta Ray Foraging Optimization (MRFO) algorithm is difficult to solve the problem of spectrum allocation, a Discrete Manta Ray Foraging Optimization (DMRFO) algorithm was proposed.Considering the pro-1 characteristic of spectrum allocation problem in engineering, firstly, MRFO algorithm was discretely binarized based on the Sigmoid Function (SF) discrete method. Secondly, the XOR operator and velocity adjustment factor were used to guide the manta rays to adaptively adjust the position of next time to the optimal solution according to the current velocity. Then, the binary spiral foraging was carried out near the global optimal solution to avoid the algorithm from falling into the local optimum. Finally, the proposed DMRFO algorithm was applied to solve the spectrum allocation problem. Simulation results show that the convergence mean and standard deviation of the network benefit when using DMRFO algorithm to allocate spectrum are 362.60 and 4.14 respectively, which are significantly better than those of Discrete Artificial Bee Colony (DABC) algorithm, Binary Particle Swarm Optimization (BPSO) algorithm and Improved Binary Particle Swarm Optimization (IBPSO) algorithm.

    Computer software technology
    Process modeling recommendation method based on behavioral profile definition target rules
    Duoqin LI, Xianwen FANG
    2022, 42(1):  223-229.  DOI: 10.11772/j.issn.1001-9081.2021010097
    Asbtract ( )   HTML ( )   PDF (627KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    In order to break the limitation of the path and graph structure in process repository based process modeling recommendation method, extract more useful recommendation information from a process repository for modelers, and assist modelers in building a business process model with higher quality, a process modeling recommendation method based on behavioral profile definition target rules was proposed. Firstly, a target profile matrix for formalizing and abstracting business interaction rules was developed through business presentation. Then, by comparing all the behavioral profile matrices in the behavioral profile matrix set with the target profile matrix, the processes in the process repository that satisfy the target profile matrix were identified as candidate process set. Finally, the process with the highest similarity to the current modeling model in the candidate process repository was calculated by the behavioral profile metric method, and the next node of the current modeling node in these processes was selected as the recommendation node. The proposed method was evaluated on a real dataset. The evaluation of both recommendation ability and recommendation accuracy shows that compared with the independent path matching method, the proposed method can provide more useful recommendation information for modelers while meeting the practical application requirements in terms of accuracy.

    Multimedia computing and computer simulation
    Image segmentation algorithm with adaptive attention mechanism based on Deeplab V3 Plus
    Zhen YANG, Xiaobao PENG, Qiangqiang ZHU, Zhijian YIN
    2022, 42(1):  230-238.  DOI: 10.11772/j.issn.1001-9081.2021010137
    Asbtract ( )   HTML ( )   PDF (1160KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    In order to solve the problem that image details and small target information are lost prematurely in the subsampling operations of Deeplab V3 Plus, an adaptive attention mechanism image semantic segmentation algorithm based on Deeplab V3 Plus network architecture was proposed. Firstly, attention mechanism modules were embedded in the input layer, middle layer and output layer of Deeplab V3 Plus backbone network, and a weight value was introduced to be multiplied with each attention mechanism module to achieve the purpose of constraining the attention mechanism modules. Secondly, the Deeplab V3 Plus embedded with the attention modules was trained on the PASCAL VOC2012 common segmentation dataset to obtain the weight values (empirical values) of the attention mechanism modules manually. Then, various fusion methods of attention mechanism modules in the input layer, the middle layer and the output layer were explored. Finally, the weight value of the attention mechanism module was automatically updated by back propagation and the optimal weight value and optimal segmentation model of the attention mechanism module were obtained. Experimental results show that, compared with the original Deeplab V3 Plus network structure, the Deeplab V3 Plus network structure with adaptive attention mechanism has the Mean Intersection over Union (MIOU) increased by 1.4 percentage points and 0.7 percentage points on the PASCAL VOC2012 common segmentation dataset and the plant pest dataset, respectively.

    Image super-resolution restoration algorithm based on information distillation network with dual attention mechanism
    Suyu WANG, Jing YANG, Yue LI
    2022, 42(1):  239-244.  DOI: 10.11772/j.issn.1001-9081.2021010134
    Asbtract ( )   HTML ( )   PDF (632KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the problems of network training difficulty and low utilization rate of feature information caused by increasing network layers in super-resolution restoration technology, an image super-resolution restoration algorithm based on dual attention Information Distillation Network (IDN) was designed and implemented. Firstly, by taking the advantage of the low computational complexity of IDN and the advantage of the information distillation module by which more features were extracted, the weights of the features were readjust adaptively by introducing the Residual Attention Module (RAM) and considering the interdependence of image channels, so as to further improve the reconstruction ability of high-resolution details of images. Then, a new mixed loss function sensitive to edge information was designed to refine the image and accelerate the convergence of the network. Test results on Set5, Set14, BSD100 and Urban100 public datasets show that the visual effect and Peak Signal-to-Noise Ratio (PSNR) of the proposed method are superior to those of the current mainstream algorithms.

    Color image demosaicking network based on inter-channel correlation and enhanced information distillation
    Hengxin LI, Kan CHANG, Yufei TAN, Mingyang LING, Tuanfa QIN
    2022, 42(1):  245-251.  DOI: 10.11772/j.issn.1001-9081.2021010127
    Asbtract ( )   HTML ( )   PDF (1841KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    In commercial digital cameras, due to the limitation of Complementary Metal Oxide Semiconductor (CMOS) sensors, there is only one color channel information for each pixel in the sampled image. Therefore, the Color image DeMosaicking (CDM) algorithm is required to restore the full-color images. However, most of the existing Convolutional Neural Network (CNN)-based CDM algorithms cannot achieve satisfactory performance with relatively low computational complexity and small network parameter number. To solve this problem, a CDM network based on Inter-channel Correlation and Enhanced Information Distillation (ICEID) was proposed. Firstly, to fully utilize the inter-channel correlation of the color image, an inter-channel guided reconstruction structure was designed to obtain the initial CDM result. Secondly, an Enhanced Information Distillation Module (EIDM), which can effectively extract and refine features from image with relatively small parameter number, was presented to enhance the reconstructed full-color image in high efficiency. Experimental results demonstrate that compared with many state-of-the-art CDM methods, the proposed algorithm achieves significant improvement in both objective quality and subjective quality, and has relatively low computational complexity and small network parameter number.

    Starting and walking human-like control of semi-passive bipedal robot with variable length telescopic legs
    Rui ZHANG, Qizhi ZHANG, Yali ZHOU
    2022, 42(1):  252-257.  DOI: 10.11772/j.issn.1001-9081.2021010175
    Asbtract ( )   HTML ( )   PDF (714KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Traditional bipedal robot walking is controlled by trajectory tracking, while human walking is in the passive state in most of the time. Aiming at the problem that the semi-passive bipedal robot with variable length telescopic legs starts to walk from a static condition, a starting and walking human-like control method was proposed. Firstly, a serial elasticity driven Bipedal Spring-Loaded Inverted Pendulum (B-SLIP) model was used. Then, the Lagrange method was used to establish the walking dynamics equation. With the self-stability of the proposed model, in the double support stage, the energy error Proportional-Integral (PI) feedback control and lazy control method were used to control the hind leg extension and contraction. In the single support stage, the swing-leg swing back method was used to control the height and forward speed of the robot. Simulation results show that the proposed control strategy can enable the bipedal robot to realize the starting and walking process on the horizontal plane, and the corresponding control system has anti-interference ability against external period disturbance force.

    Frontier and comprehensive applications
    Improved spatio-temporal residual convolutional neural network for urban road network short-term traffic flow prediction
    Yinxin BAO, Yang CAO, Quan SHI
    2022, 42(1):  258-264.  DOI: 10.11772/j.issn.1001-9081.2021010080
    Asbtract ( )   HTML ( )   PDF (1139KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Traffic flow prediction for urban road network is influenced by historical traffic flow and traffic flow at adjacent intersections, which has complex spatio-temporal correlation. For the lack of correlation analysis on traffic flow data, capturing small changes but ignoring long-term time characteristics in traditional spatio-temporal residual models, a short-term traffic flow prediction model for urban road network based on improved spatio-temporal residual Convolutional Neural Network (CNN) was proposed. In the proposed model, the original traffic flow data was transformed into traffic grid data, and Pearson Correlation Coefficient (PCC) was used to analyze the correlation of traffic grid data, so as to determine the periodic series and adjacent series with high correlation. At the same time, the periodic series model and the adjacent series model were established, and Long Short-Term Memory (LSTM) network was introduced as the hybrid model to extract the time characteristics and capture the long-term time characteristics of the two series. Experimental results on Chengdu taxi dataset show that the proposed model can predict traffic flow better than benchmark models of LSTM, CNN and the traditional residual model. When the evaluation index is Root Mean Square Error (RMSE), the average prediction accuracy of traffic road network in the test set is improved by 25.6%, 13.3% and 3.2% respectively.

    Elongated pavement distress detection method based on convolutional neural network
    Huiqing XU, Bin CHEN, Jingfei WANG, Zhiyi CHEN, Jian QIN
    2022, 42(1):  265-272.  DOI: 10.11772/j.issn.1001-9081.2021010206
    Asbtract ( )   HTML ( )   PDF (2146KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Focusing on the problems of the large time consumption of manual detection and the insufficient precision of the current detection methods of elongated pavement distress, a two-stage elongated pavement distress detection method, named Epd RCNN (Elongated pavement distress Region-based Convolutional Neural Network), which could accurately locate and classify the distress was proposed according to the weak semantic characteristics and abnormal geometric properties of the distress. Firstly, for the weak semantic characteristics of elongated pavement distress, a backbone network that reused low-level features and repeatedly fused the features of different stages was proposed. Secondly, in the training process, the high-quality positive samples for network training were generated by the anchor box mechanism conforming to the geometric property distribution of the distress. Then, the distress bounding boxes were predicted on a single high-resolution feature map, and a parallel cascaded dilated convolution module was used to this feature map to improve its multi-scale feature representation ability. Finally, for different shapes of region proposals, the region proposal features conforming to the distress geometric properties were extracted by the proposal feature improvement module composed of deformable Region of Interest Pooling (RoI Pooling) and spatial attention module. Experimental results show that the proposed method has the mean Average Precision (mAP) of 0.907 on images with sufficient illumination, the mAP of 0.891 on images with illumination problems and the comprehensive mAP of 0.899, indicating that the proposed method has good detection performance and robustness to illumination.

    Valve identification method based on double detection
    Wei SHE, Qian ZHENG, Zhao TIAN, Wei LIU, Yinghao LI
    2022, 42(1):  273-279.  DOI: 10.11772/j.issn.1001-9081.2021020333
    Asbtract ( )   HTML ( )   PDF (2307KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the problems that current valve identification methods in industry have high missed rate of overlapping targets, low detection precision, poor target encapsulation degree and inaccurate positioning of circle center, a valve identification method based on double detection was proposed. Firstly, data enhancement was used to expand the samples in a lightweight way. Then, Spatial Pyramid Pooling (SPP) and Path Aggregation Network (PAN) were added on the basis of deep convolutional network. At the same time, the anchor boxes were adjusted and the loss function was improved to extract the valve prediction boxes. Finally, the Circle Hough Transform (CHT) method was used to secondarily identify the valves in the prediction boxes to accurately identify the valve regions. The proposed method was compared with the original You Only Look Once (YOLO)v3, YOLOv4, and the traditional CHT methods, and the detection results were evaluated by jointly using precision, recall and coincidence degree. Experimental results show that the average precision and recall of the proposed method reaches 97.1% and 94.4% respectively, 2.9 percentage points and 1.8 percentage points higher than those of the original YOLOv3 method respectively. In addition, the proposed method improves the target encapsulation degree and location accuracy of target center. The proposed method has the Intersection Over Union (IOU) between the corrected frame and the real frame reached 0.95, which is 0.05 higher than that of the traditional CHT method. The proposed method improves the success rate of target capture while improving the accuracy of model identification, and has certain practical value in practical applications.

    Application of Stacking-Bagging-Vote multi-source information fusion model for financial early warning
    Lu ZHANG, Jiapeng LIU, Dongmei TIAN
    2022, 42(1):  280-286.  DOI: 10.11772/j.issn.1001-9081.2021020306
    Asbtract ( )   HTML ( )   PDF (948KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Ensemble resampling technology can solve the problem of imbalanced samples in financial early warning research to some extent. Different ensemble models and different ensemble resampling technologies have different suitabilities. It is found in the study that Up-Down ensemble sampling and Tomek-Smote ensemble sampling were respectively suitable for Bagging-Vote ensemble model and Stacking fusion model. Based on the above, a Stacking-Bagging-Vote (SBV) multi-source information fusion model was built. Firstly, the Bagging-Vote model based on Up-Down ensemble sampling and the Stacking model based on Tomek-Smote sampling were fused. Then, the stock trading data were added and processed by Kalman filtering, so that the interactive fusion optimization of data level and model level was realized, and the SBV multi-source information fusion model was finally obtained. This fusion model not only has a great improvement in the prediction performance by taking into account prediction accuracy and prediction precision simultaneously, but also can select the corresponding SBV multi-source information fusion model to perform the financial early warning to meet the actual needs of different stakeholders by adjusting the parameters of the model.

    Multi-site temperature prediction model based on graph convolutional network and gated recurrent unit
    Donglin MA, Sizhou MA, Weijie WANG
    2022, 42(1):  287-293.  DOI: 10.11772/j.issn.1001-9081.2021010099
    Asbtract ( )   HTML ( )   PDF (753KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Spatio-temporal prediction task is widely applied in neuroscience, transportation, meteorology and other fields. As a typical spatio-temporal prediction task, temperature prediction needs to dig out the inherent spatio-temporal characteristics of temperature data. Aiming at the problems of large prediction error and insufficient spatial feature extraction in the existing temperature prediction algorithms, a temperature prediction model based on Graph Convolutional Network and Gated Recurrent Unit (GCN-GRU) was proposed. Firstly, the methods of weight redistribution and multi-order neighbor connection were used to modify Graph Convolutional Network (GCN) in order to effectively mine the unique spatial characteristics of the meteorological data. Secondly, the matrix multiplication of each recurrent unit in the Gated Recurrent Unit (GRU) was replaced by graph convolution operation, and all the recurrent units were connected in series to form a graph convolutional gating layer. Then, the graph convolutional gating layer was used to build the main network structure to extract the spatio-temporal characteristics of the data. Finally, the temperature prediction results were output through a fully connected output layer. Compared with the single models such as Gated Recurrent Unit (GRU) and Long Short-Term Memory (LSTM), GCN-GRU had the Mean Absolute Error (MAE)reduced by 0.67 and 0.83 respectively; compared with the prediction model combined with Chebyshev graph convolution and Long Short-Term Memory (Cheb-LSTM) and the prediction model combined with Graph Convolutional Network and Long Short-Term Memory (GCN-LSTM), the proposed model had the MAE reduced by 0.36 and 0.23 respectively.

    Fall behavior detection algorithm for the elderly based on AlphaPose optimization model
    Jingqi MA, Huan LEI, Minyi CHEN
    2022, 42(1):  294-301.  DOI: 10.11772/j.issn.1001-9081.2021020331
    Asbtract ( )   HTML ( )   PDF (7482KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    In order to detect the elderly fall high-risk behaviors quickly and accurately on the low-power and low-cost hardware platform, an abnormal behavior detection algorithm based on AlphaPose optimization model was proposed. Firstly, the pedestrian target detection model and pose estimation model were optimized to accelerate the human target detection and pose joint point reasoning. Then, the image coordinate data of human pose joint points were computed rapidly through the optimized AlphaPose model. Finally, the relationship between the head joint point linear velocity and the crotch joint linear velocity at the moment the human body falls was calculated, as well as the change of the angle between the midperpendicular of the torso and X-axis of the image, were calculated to determine the occurrence of the fall. The proposed algorithm was deployed to the Jetson Nano embedded development board, and compared with several main fall detection algorithms based on human pose at present: YOLO (You Only Look Once)v3+Pose, YOLOv4+Pose, YOLOv5+Pose, trt_pose and NanoDet+Pose. Experimental results show that on the used embedded platform when the image resolution is 320×240, the proposed algorithm has the detection frame rate of 8.83 frame/s and the accuracy of 0.913, which are both better than those of the algorithms compared above. The proposed algorithm has relatively high real-time performance and accuracy, and can timely detect the occurrence of the elderly fall behaviors.

    Weakly supervised fine-grained classification method of Alzheimer’s disease based on improved visual geometry group network
    Shuang DENG, Xiaohai HE, Linbo QING, Honggang CHEN, Qizhi TENG
    2022, 42(1):  302-309.  DOI: 10.11772/j.issn.1001-9081.2021020258
    Asbtract ( )   HTML ( )   PDF (868KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    In order to solve the problems of small difference of Magnetic Resonance Imaging (MRI) images between Alzheimer’s Disease (AD) patients and Normal Control (NC) people and great difficulty in classification of them, a weakly supervised fine-grained classification method for AD based on improved Visual Geometry Group (VGG) network was proposed. In this method, Weakly Supervised Data Augmentation Network (WSDAN) was took as the basic model, which was mainly composed of weakly supervised attention learning module, data augmentation module and bilinear attention pooling module. Firstly, the feature map and the attention map were generated through weakly supervised attention learning network, and the attention map was used to guide the data augmentation. Both the original image and the augmented data were used as the input data for training. Then, point production between the feature map and the attention map was performed by elements via bilinear attention pooling algorithm to obtain the feature matrix. Finally, the feature matrix was used as the input of the linear classification layer. Experimental results of applying WSDAN basic model with VGG19 as feature extraction network on MRI data of AD show that, compared with the WSDAN basic model, the proposed model only with image enhancement has the accuracy, sensitivity and specificity increased by 1.6 percentage points, 0.34 percentage points and 0.12 percentage points respectively; the model only using the improvement of VGG19 network has the accuracy and specificity improved by 0.7 percentage points and 2.82 percentage points respectively; the model combing the two methods above has the accuracy, sensitivity and specificity improved by 2.1 percentage points, 1.91 percentage points and 2.19 percentage points respectively.

    Application of 3DPCANet in image classification of functional magnetic resonance imaging for Alzheimer’s disease
    Hongfei JIA, Xi LIU, Yu WANG, Hongbing XIAO, Suxia XING
    2022, 42(1):  310-315.  DOI: 10.11772/j.issn.1001-9081.2021010132
    Asbtract ( )   HTML ( )   PDF (568KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Alzheimer’s Disease (AD) is a progressive neurodegenerative disease with hidden causes, and can result in structural changes of patients’ brain regions. For assisting the doctors to make correct judgment on the condition of AD patients, an improved Three-Dimensional Principal Component Analysis Network (3DPCANet) model was proposed to classify AD by combining the mean Amplitude of Low-Frequency Fluctuation (mALFF) image of the whole brain of the subject. Firstly, functional Magnetic Resonance Imaging (fMRI) data were preprocessed, and the mALFF image of the whole brain was calculated. Then, the improved 3DPCANet deep learning model was used for feature extraction. Finally, Support Vector Machine (SVM) was used to classify features of AD patients with different stages. Experimental results show that the proposed model is simple and robust, and has the classification accuracies on Subjective Memory Decline (SMD) vs. AD, SMD vs. Late Mild Cognitive Impairment (LMCI), and LMCI vs. AD reached 92.42%, 91.80% and 89.33% respectively, which verifies the effectiveness and feasibility of the proposed method.

    Semi-supervised knee abnormality classification based on multi-imaging center MRI data
    Jie WU, Shitian ZHANG, Haibin XIE, Guang YANG
    2022, 42(1):  316-324.  DOI: 10.11772/j.issn.1001-9081.2021010200
    Asbtract ( )   HTML ( )   PDF (780KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The manual labeling of abundant data is laborious and the amount of Magnetic Resonance Imaging (MRI) data from a single imaging center is limited. Concerning the above problems, a Magnetic Resonance Semi-Supervised Learning (MRSSL) method utilizing multi-imaging center labeled and unlabeled MRI data was proposed and applied to knee abnormality classification. Firstly, data augmentation was used to provide the inductive bias required by the model . Next, the classification loss and the consistency loss were combined to constraint an artificial neural network to extract the discriminative features from the data. Then, the features were used for the MRI knee abnormality classification. Additionally, the corresponding Magnetic Resonance Supervised Learning (MRSL) method only using labeled samples was proposed and compared with MRSSL for the same labeled samples. The results demonstrate that MRSSL surpasses MRSL in both model classification performance and model generalization ability. Finally, MRSSL was compared with other semi-supervised learning methods. The results indicate that data augmentation plays an important role on performance improvement, and with stronger inclusiveness for MRI data, MRSSL outperforms others on the knee abnormality classification.

    Multi-aspect multi-attention fusion of molecular features for drug-target affinity prediction
    Runze WANG, Yueqin ZHANG, Qiqi QIN, Zehua ZHANG, Xumin GUO
    2022, 42(1):  325-332.  DOI: 10.11772/j.issn.1001-9081.2021071218
    Asbtract ( )   HTML ( )   PDF (1149KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Recent deep learning achieves great attention on the tasks of Drug-Target Affinity (DTA). However, most existing works embed the molecular single structure as a vector, while ignoring the information gain provided by multi-aspect fusion of molecular features to the final feature representation. To address the feature incompleteness problem of single-structured molecules, an end-to-end deep learning method based on attentive fusion of multi-aspect molecular features was proposed for DTA prediction. Multi-aspect molecular structure embedding (Mas) and Multi-attention feature fusion (Mat) are the core modules of the proposed method. Firstly, the multi-aspect molecular structure was embedded into the feature vector space by Mas module. Secondly, the attention mechanism of molecular feature level was incorporated for the weighted fusion of molecular features from different aspects through Mat module. Thirdly, feature cascade of the above two was performed according to Drug-Target Interaction (DTI). Finally, the fully connected neural network was used to realize the regression prediction of the affinity. Experiments on Davis and KIBA datasets were carried out to evaluate the influence of training ratio, multi-aspect feature incorporation, multi-attention fusion, and related parameters on the performance of affinity prediction. Compared with the GraphDTA method, the proposed method has the Mean Square Error (MSE) reduced by 4.8% and 6% on the two datasets, respectively. Experimental results show that attentive fusion of multi-aspect molecular features can capture the molecular features that are more relevant for linkages on protein targets.

2024 Vol.44 No.3

Current Issue
Archive
Honorary Editor-in-Chief: ZHANG Jingzhong
Editor-in-Chief: XU Zongben
Associate Editor: SHEN Hengtao XIA Zhaohui
Domestic Post Distribution Code: 62-110
Foreign Distribution Code: M4616
Address:
No. 9, 4th Section of South Renmin Road, Chengdu 610041, China
Tel: 028-85224283-803
  028-85222239-803
Website: www.joca.cn
E-mail: bjb@joca.cn
WeChat
Join CCF