Loading...

Table of Content

    10 December 2023, Volume 43 Issue 12
    Artificial intelligence
    Gradient descent with momentum algorithm based on differential privacy in convolutional neural network
    Yu ZHANG, Ying CAI, Jianyang CUI, Meng ZHANG, Yanfang FAN
    2023, 43(12):  3647-3653.  DOI: 10.11772/j.issn.1001-9081.2022121881
    Asbtract ( )   HTML ( )   PDF (1985KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    To address the privacy leakage problem caused by the model parameters memorizing some features of the data during the training process of the Convolutional Neural Network (CNN) models, a Gradient Descent with Momentum algorithm based on Differential Privacy in CNN (DPGDM) was proposed. Firstly, the Gaussian noise meeting differential privacy was added to the gradient in the backpropagation process of model optimization, and the noise-added gradient value was used to participate in the model parameter update process, so as to achieve differential privacy protection for the overall model. Secondly, to reduce the impact of the introduction of differential privacy noise on convergence speed of the model, a learning rate decay strategy was designed and then the gradient descent with momentum algorithm was improved. Finally, to reduce the influence of noise on the accuracy of the model, the value of the noise scale was adjusted dynamically during model optimization, thereby changing the amount of noise that needs to be added to the gradient in each round of iteration. Experimental results show that compared with DP-SGD (Differentially Private Stochastic Gradient Descent) algorithm, the proposed algorithm can improve the accuracy of the model by about 5 and 4 percentage points at privacy budget of 0.3 and 0.5, respectively, proving that by using the proposed algorithm, the model usability is improved and privacy protection of the model is achieved.

    Heterogeneous hypernetwork representation learning method with hyperedge constraint
    Keke WANG, Yu ZHU, Xiaoying WANG, Jianqiang HUANG, Tengfei CAO
    2023, 43(12):  3654-3661.  DOI: 10.11772/j.issn.1001-9081.2022121908
    Asbtract ( )   HTML ( )   PDF (2264KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Compared with ordinary networks, hypernetworks have complex tuple relationships, namely hyperedges. However, most existing network representation learning methods cannot capture the tuple relationships. To solve the above problem, a Heterogeneous hypernetwork Representation learning method with Hyperedge Constraint (HRHC) was proposed. Firstly, a method combining clique extension and star extension was introduced to transform the heterogeneous hypernetwork into the heterogeneous network. Then, the meta-path walk method that was aware of semantic relevance among the nodes was introduced to capture the semantic relationships among the heterogeneous nodes. Finally, the tuple relationships among the nodes were captured by means of the hyperedge constraint to obtain high-quality node representation vectors. Experimental results on three real-world datasets show that, for the link prediction task, the proposed method obtaines good results on drug, GPS and MovieLens datasets. For the hypernetwork reconstruction task, when the hyperedge reconstruction ratio is more than 0.6, the ACCuracy (ACC) of the proposed method is better than the suboptimal method Hyper2vec(biased 2nd order random walks in Hyper-networks), and the average ACC of the proposed method outperforms the suboptimal method, that is heterogeneous hypernetwork representation learning method with hyperedge constraint based on incidence graph (HRHC-incidence graph) by 15.6 percentage points on GPS dataset.

    Deep spectral clustering algorithm with L1 regularization
    Wenbo LI, Bo LIU, Lingling TAO, Fen LUO, Hang ZHANG
    2023, 43(12):  3662-3667.  DOI: 10.11772/j.issn.1001-9081.2022121822
    Asbtract ( )   HTML ( )   PDF (1465KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the problems that the deep spectral clustering models perform poorly in training stability and generalization capability, a Deep Spectral Clustering algorithm with L1 Regularization (DSCLR) was proposed. Firstly, L1 regularization was introduced into the objective function of deep spectral clustering to sparsify the eigen vectors of the Laplacian matrix generated by the deep neural network model. And the generalization capability of the model was enhanced. Secondly, the network structure of the spectral clustering algorithm based on deep neural network was improved by using the Parametric Rectified Linear Unit activation function (PReLU) to solve the problems of model training instability and underfitting. Experimental results on MNIST dataset show that the proposed algorithm improves Clustering Accuracy (CA), Normalized Mutual Information (NMI) index, and Adjusted Rand Index (ARI) by 11.85, 7.75, and 17.19 percentage points compared to the deep spectral clustering algorithm, respectively. Furthermore, the proposed algorithm also significantly improves the three evaluation metrics, CA, NMI and ARI, compared to algorithms such as Deep Embedded Clustering (DEC) and Deep Spectral Clustering using Dual Autoencoder Network (DSCDAN).

    Zero-shot relation extraction model via multi-template fusion in Prompt
    Liang XU, Chun ZHANG, Ning ZHANG, Xuetao TIAN
    2023, 43(12):  3668-3675.  DOI: 10.11772/j.issn.1001-9081.2022121869
    Asbtract ( )   HTML ( )   PDF (1768KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Prompt paradigm is widely used to zero-shot Natural Language Processing (NLP) tasks. However, the existing zero-shot Relation Extraction (RE) model based on Prompt paradigm suffers from the difficulty of constructing answer space mappings and dependence on manual template selection, which leads to suboptimal performance. To address these issues, a zero-shot RE model via multi-template fusion in Prompt was proposed. Firstly, the zero-shot RE task was defined as the Masked Language Model (MLM) task, where the construction of answer space mapping was abandoned. Instead, the words output by the template were compared with the relation description text in the word embedding space to determine the relation class. Then, the part of speech of the relation description text was introduced as a feature, and the weight between this feature and each template was learned. Finally, this weight was utilized to fuse the results output by multiple templates, thereby reducing the performance loss caused by the manual selection of Prompt templates. Experimental results on FewRel (Few-shot Relation extraction dataset) and TACRED (Text Analysis Conference Relation Extraction Dataset) show that, the proposed model significantly outperforms the current state-of-the-art model, RelationPrompt, in terms of F1 score under different data resource settings, with an increase of 1.48 to 19.84 percentage points and 15.27 to 15.75 percentage points, respectively. These results convincingly demonstrate the effectiveness of the proposed model for zero-shot RE tasks.

    Hierarchical algorithm of fuzzy Petri net by reverse search
    Yinhong XIANG, Kaiqing ZHOU, Senyu YANG, Xuanyu ZHANG, Diwen KANG
    2023, 43(12):  3676-3682.  DOI: 10.11772/j.issn.1001-9081.2022121851
    Asbtract ( )   HTML ( )   PDF (2123KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Fuzzy Petri Net (FPN) is one of the main tools to represent, model, and analyze the Knowledge-Based System (KBS). For clear hierarchical structures and uncertain affiliations between places/transactions in some FPNs, a Hierarchical algorithm of FPN by Reverse Search (HFPN-RS) was proposed to realize the automatic conversion from a non-hierarchical FPN to a Hierarchical FPN (HFPN). Firstly, a reverse search of the entire FPN was launched starting from the output place(s) at first. The front set of input place(s) and the back set of output place(s) were divided into the same layer. Then, the hierarchical structure of the entire FPN was clarified by adding virtual place-virtual transition pairs. Meanwhile, two theorems were proven to define the infimum of the number of hierarchical layers of FPN and the minimum number of virtual place-virtual transition pair(s) that need to be added in the hierarchical operation, respectively. Moreover, the dimension calculation formula of the incidence matrix of the complete hierarchical structure was also introduced. In the experimental part, hierarchical operation was performed on several types of FPN models with different characteristics and the proposed theorems were used to verify the HFPN-RS algorithm. The experimental results show that the new FPN has a clear hierarchical structure by adding the virtual place-virtual transition pair(s). It provides a theoretical base to further study the FPN generalization ability.

    Contrastive hypergraph transformer for session-based recommendation
    Weichao DANG, Bingyang CHENG, Gaimei GAO, Chunxia LIU
    2023, 43(12):  3683-3688.  DOI: 10.11772/j.issn.1001-9081.2022111654
    Asbtract ( )   HTML ( )   PDF (1447KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    A Contrastive Hypergraph Transformer for session-based recommendation (CHT) model was proposed to address the problems of noise interference and sample sparsity in the session-based recommendation itself. Firstly, the session sequence was modeled as a hypergraph. Secondly, the global context information and local context information of items were constructed by the hypergraph transformer. Finally, the Item-Level (I-L) encoder and Session-Level (S-L) encoder were used on global relationship learning to capture different levels of item embeddings, the information fusion module was used to fuse item embedding and reverse position embedding, and the global session representation was obtained by the soft attention module while the local session representation was generated with the help of the weight line graph convolutional network on local relationship learning. In addition, a contrastive learning paradigm was introduced to maximize the mutual information between the global and local session representations to improve the recommendation performance. Experimental results on several real datasets show that the recommendation performance of CHT model is better than that of the current mainstream models. Compared with the suboptimal model S2-DHCN (Self-Supervised Hypergraph Convolutional Networks), the proposed model has the P@20 of 35.61% and MRR@20 of 17.11% on Tmall dataset, which are improved by 13.34% and 13.69% respectively; the P@20 reached 54.07% and MRR@20 reached 18.59% on Diginetica dataset, which are improved by 0.76% and 0.43% respectively; verifying the effectiveness of the proposed model.

    Session-based recommendation model by graph neural network fused with item influence
    Xuanyu SUN, Yancui SHI
    2023, 43(12):  3689-3696.  DOI: 10.11772/j.issn.1001-9081.2022121812
    Asbtract ( )   HTML ( )   PDF (2062KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the problem that it is difficult for the existing session-based recommendation models to explicitly express the influence of items on the recommendation results, a Session-based Recommendation model by graph neural network fused with Item Influence (SR-II) was proposed. Firstly, a new edge weight calculation method was proposed to construct a graph structure, in which the calculated result was used as the influence weight of the transition relationship in the graph, and the features of the graph were extracted through the influence graph gated layer by using Graph Neural Network (GNN). Then, an improved shortcut graph was proposed to connect related items, effectively capture long-range dependencies, and enrich the information expressed by the graph structure; and the features of the graph were extracted through the shortcut graph attention layer by using the attention mechanism. Finally, a recommendation model was constructed by combining the above two layers. In the experimental results on Diginetica and Gowalla datasets, the highest HR@20 of SR-II is reaching 53.12%, and the highest MRR@20 of SR-II is reaching 25.79%. On Diginetica dataset, compared with CORE-trm (simple and effective session-based recommendation within COnsistent REpresentation space-transformer), SR-II has the HR@20 improved by 1.10% ,and the MRR@20 improved by 1.21%; On Gowalla dataset, compared with SR-SAN(Session-based Recommendation with Self-Attention Networks), SR-II has the HR@20 improved by 1.73%.Compared with the recommendation model called LESSR (Lossless Edge-order preserving aggregation and Shortcut graph attention for Session-based Recommendation), SR-II has the MRR@20 improved by 1.14%. The experimental results show that the performance of SR-II is better than that of the comparison models, and SR-II has a higher recommendation accuracy.

    Knowledge concept recommendation system based on interest enhancement
    Yu LING, Zhilong SHAN
    2023, 43(12):  3697-3702.  DOI: 10.11772/j.issn.1001-9081.2022111786
    Asbtract ( )   HTML ( )   PDF (1346KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The existing knowledge concept recommendation system does not consider the short-term interest of users. To solve the problem, a Knowledge Concept Recommendation system based on Interest Enhancement (KCRec-IE) was proposed. Firstly, users’ short-term interests were captured according to the users’ knowledge concept click sequences, and a heterogeneous graph was constructed by using the side information. Then, the representation learning of knowledge concept entities and user entities was carried out on heterogeneous graph by using meta-path-guided graph convolution. Different from the representation learning of knowledge concept entities, when learning the representation of user entities, the contributions of different neighbor users to target users were able to be distinguished according to the short-term interests of users. Finally, the score prediction was realized according to the knowledge concept entities, the user entities and the user’s short-term interests. Experimental results on public dataset Xuetang X show that compared with KCRec-SEIGNN, KCRec-IE is improved by 3.60 percentage points on HR@5; compared with KCRec-IEn, KCRec-IE is improved by 1.02 percentage points on HR@10; compared with KCRec-SEIGNN, KCRec-IE is improved by 1.60 and 1.18 percentage points respectively on NDGC@5 and NDGC@10 respectively, verifying the effectiveness of the proposed method.

    Semantically enhanced sentiment classification model based on multi-level attention
    Jianle CAO, Nana LI
    2023, 43(12):  3703-3710.  DOI: 10.11772/j.issn.1001-9081.2022121894
    Asbtract ( )   HTML ( )   PDF (1638KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The existing text sentiment classification methods face serious challenges due to the complex semantics of natural language, the multiple sentiment polarities of words, and the long-term dependency of text. To solve these problems, a semantically enhanced sentiment classification model based on multi-level attention was proposed. Firstly, the contextualized dynamic word embedding technology was used to mine the multiple semantic information of words, and the context semantics was modeled. Secondly, the long-term dependency within the text was captured by the multi-layer parallel multi-head self-attention in the internal attention layer to obtain comprehensive text feature information. Thirdly, in the external attention layer, the summary information in the review metadata was integrated into the review features through a multi-level attention mechanism to enhance the sentiment information and semantic expression ability of the review features. Finally, the global average pooling layer and Softmax function were used to realize sentiment classification. Experimental results on four Amazon review datasets show that, compared with the best-performing TE-GRU (Transformer Encoder with Gated Recurrent Unit) in the baseline models, the proposed model improves the sentiment classification accuracy on App, Kindle, Electronic and CD datasets by at least 0.36, 0.34, 0.58 and 0.66 percentage points, which verifies that the proposed model can further improve the sentiment classification performance.

    Chinese word segmentation method in electric power domain based on improved BERT
    Fei XIA, Shuaiqi CHEN, Min HUA, Bihong JIANG
    2023, 43(12):  3711-3718.  DOI: 10.11772/j.issn.1001-9081.2022121897
    Asbtract ( )   HTML ( )   PDF (1953KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    To solve the problem of poor performance in segmenting a large number of proprietary words in Chinese text in electric power domain, an improved Chinese Word Segmentation (CWS) method in electric power domain based on improved BERT (Bidirectional Encoder Representations from Transformer) was proposed. Firstly, two lexicons were built covering general words and domain words respectively, and a dual-lexicon matching and integration mechanism was designed to directly integrate the word features into BERT model, enabling more effective utilization of external knowledge by the model. Then, DEEPNORM method was introduced to improve the model’s ability to extract features, and the optimal depth of the model was determined by Bayesian Information Criterion (BIC), which made BERT model stable up to 40 layers. Finally, the classical self-attention layer in BERT model was replaced by the ProbSparse self-attention layer, and the best value of sampling factor was determined by using Particle Swarm Optimization (PSO) algorithm to reduce the model complexity while ensuring the model performance. The test of word segmentation was carried out on a hand-labeled patent text dataset in electric power domain. Experimental results show that the proposed method achieves the F1 score of 92.87%, which is 14.70, 9.89 and 3.60 percentage points higher than those of the methods to be compared such as Hidden Markov Model (HMM), multi-standard word segmentation model METASEG (pre-training model with META learning for Chinese word SEGmentation) and Lexicon Enhanced BERT (LEBERT) model, verifying that the proposed method effectively improves the quality of Chinese text word segmentation in electric power domain.

    Cloth-changing person re-identification model based on semantic-guided self-attention network
    Jianhua ZHONG, Chuangyi QIU, Jianshu CHAO, Ruicheng MING, Jianfeng ZHONG
    2023, 43(12):  3719-3726.  DOI: 10.11772/j.issn.1001-9081.2022121875
    Asbtract ( )   HTML ( )   PDF (2046KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Focused on the difficulty of extracting effective information in the cloth-changing person Re-identification (ReID) task, a cloth-changing person re-identification model based on semantic-guided self-attention network was proposed. Firstly, semantic information was used to segment an original image into a cloth-free image. Both images were input into a two-branch multi-head self-attention network to extract cloth-independent features and complete person features, respectively. Then, a Global Feature Reconstruction module (GFR) was designed to reconstruct two global features, in which the clothing region contained head features with better robustness, which made the saliency information in the global features more prominent. And a Local Feature Reorganization and Reconstruction module (LFRR) was proposed to extract the head and shoe features from the original image and the cloth-free image, emphasizing the detailed information about the head and shoe features and reducing the interference caused by changing shoes. Finally, in addition to the identity loss and triplet loss commonly used in person re-identification, Feature Pull Loss (FPL) was proposed to close the distances among local and global features, complete image features and costume-free image features. On the PRCC (Person ReID under moderate Clothing Change) and VC-Clothes (Virtually Changing-Clothes) datasets, the mean Average Precision (mAP) of the proposed model improved by 4.6 and 0.9 percentage points respectively compared to the Clothing-based Adversarial Loss (CAL) model. On the Celeb-reID (Celebrities re-IDentification) and Celeb-reID-light (a light version of Celebrities re-IDentification) datasets, the mAP of the proposed model improved by 0.2 and 5.0 percentage points respectively compared with the Joint Loss Capsule Network (JLCN) model. The experimental results show that the proposed method has certain advantages in highlighting effective information expression in the cloth-changing scenarios.

    Text-independent speaker verification method based on uncertainty learning
    Yulian ZHANG, Shanshan YAO, Chao WANG, Jiang CHANG
    2023, 43(12):  3727-3732.  DOI: 10.11772/j.issn.1001-9081.2022121902
    Asbtract ( )   HTML ( )   PDF (1147KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The speaker verification task aims to determine whether a registration speech and a test speech belong to the same speaker. A Text-Independent Speaker Verification (TISV) method based on Uncertainty Learning (UL) was proposed to address the problem that the voiceprint features extracted by speaker recognition systems are usually disturbed by factors unrelated to identity information, thereby leading to serious degradation of the system accuracy. Firstly, uncertainty was introduced in the speaker backbone network to simultaneously learn the voiceprint features (mean) and the uncertainty of the speech data (variance), so as to model the uncertainty in the speech dataset. Then, the distribution representation of the features was obtained by a resampling technique. Finally, the degradation problem in the calculation process of classification loss was solved by constraining the distribution of the noise through the introduction of KL (Kullback-Leibler) divergence regularization into the speaker classification loss. Experimental results show that after training on VoxCeleb1 and VoxCeleb2 development sets and testing on VoxCeleb1-O test set, compared with the certainty method-based model Thin ResNet34, the model of the proposed method has the Equal Error Rate (EER) reduced by 9.9% and 10.4% respectively, and minimum Detection Cost Function (minDCF) reduced by 10.9% and 4.5% respectively. It can be seen that the accuracy of the proposed method is improved in noisy and unconstrained scenarios.

    SiamTrans: tiny object tracking algorithm based on Siamese network and Transformer
    Haitao GONG, Zhihua CHEN, Bin SHENG, Bingyan ZHU
    2023, 43(12):  3733-3739.  DOI: 10.11772/j.issn.1001-9081.2022111790
    Asbtract ( )   HTML ( )   PDF (2957KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the problems of poor robustness, low precision and success rate in the existing tiny object tracking algorithms, a tiny object tracking algorithm, SiamTrans, was proposed on the basis of Siamese network and Transformer. Firstly, a similarity response map calculation module was designed based on the Transformer mechanism. In the module, several layers of feature encoding-decoding structures were superimposed, and multi-head self-attention and multi-head cross-attention mechanisms were used to query template feature map information in feature maps of different levels of search regions, which avoided falling into local optimal solutions and obtained a high-quality similarity response map. Secondly, a Prediction Module (PM) based on Transformer mechanism was designed in the prediction subnetwork, and the self-attention mechanism was used to process redundant feature information in the prediction branch feature maps to improve the prediction precisions of different prediction branches. Experimental results on Small90 dataset show that, compared to the TransT (Transformer Tracking) algorithm, the tracking precision and tracking success rate of the proposed algorithm are 8.0 and 9.5 percentage points higher, respectively. It can be seen that the proposed algorithm has better tracking performance for tiny objects.

    Data science and technology
    Contrast order-preserving pattern mining algorithm
    Yufei MENG, Youxi WU, Zhen WANG, Yan LI
    2023, 43(12):  3740-3746.  DOI: 10.11772/j.issn.1001-9081.2022121828
    Asbtract ( )   HTML ( )   PDF (909KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the problem that the existing contrast sequential pattern mining methods mainly focus on character sequence datasets and are difficult to be applied to time series datasets, a new Contrast Order-preserving Pattern Mining (COPM) algorithm was proposed. Firstly, in the candidate pattern generation stage, a pattern fusion strategy was used to reduce the number of candidate patterns. Then, in the pattern support calculation stage, the support of super-pattern was calculated by using the matching results of sub-patterns. Finally, a dynamic pruning strategy of minimum support threshold was designed to further effectively prune the candidate patterns. Experimental results show that on six real time series datasets, the memory consumption of COPM algorithm is at least 52.1% lower than that of COPM-o (COPM-original) algorithm, 36.8% lower than that of COPM-e (COPM-enumeration) algorithm, and 63.6% lower than that of COPM-p (COPM-prune) algorithm. At the same time, the running time of COPM algorithm is at least 30.3% lower than that of COPM-o algorithm, 8.8% lower than that of COPM-e algorithm and 41.2% lower than that of COPM-p algorithm. Therefore, in terms of algorithm performance, COPM algorithm is superior to COPM-o, COPM-e and COPM-p algorithms. The experimental results verify that COPM algorithm can effectively mine the contrast order-preserving patterns to find the differences between different classes of time series datasets.

    Large-scale subspace clustering algorithm with Local structure learning
    Qize REN, Hongjie JIA, Dongyu CHEN
    2023, 43(12):  3747-3754.  DOI: 10.11772/j.issn.1001-9081.2022111750
    Asbtract ( )   HTML ( )   PDF (768KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The conventional large-scale subspace clustering methods ignore the local structure that prevails among the data when computing the anchor affinity matrix, and have large error when calculating the approximate eigenvectors of the Laplacian matrix, which is not conducive to data clustering. Aiming at the above problems, a Large-scale Subspace Clustering algorithm with Local structure learning (LLSC) was proposed. In the proposed algorithm, the local structure learning was embedded into the learning of anchor affinity matrix, which was able to comprehensively use global and local information to mine the subspace structure of data. In addition, inspired by Nonnegative Matrix Factorization (NMF), an iterative optimization method was designed to simplify the solution of anchor affinity matrix. Then, the mathematical relationship between the anchor affinity matrix and the Laplacian matrix was established according to the Nystr?m approximation method, and the calculation method of the eigenvectors of the Laplacian matrix was modified to improve the clustering performance. Compared to LMVSC (Large-scale Multi-View Subspace Clustering), SLSR (Scalable Least Square Regression), LSC-k (Landmark-based Spectral Clustering using k-means), and k-FSC(k-Factorization Subspace Clustering), LLSC demonstrates significant improvements on four widely used large-scale datasets. Specifically, on the Pokerhand dataset, the accuracy of LLSC is 28.18 points percentage higher than that of k-FSC. These results confirm the effectiveness of LLSC.

    Agglomerative hierarchical clustering algorithm based on hesitant fuzzy set
    Wenquan LI, Yimin MAO, Xindong PENG
    2023, 43(12):  3755-3763.  DOI: 10.11772/j.issn.1001-9081.2023010094
    Asbtract ( )   HTML ( )   PDF (626KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the problems of information distortion, poor objectivity of attribute weights, and high time complexity in hesitant fuzzy clustering analysis, an Agglomerative Hierarchical Clustering algorithm based on Hesitant Fuzzy set (AHCHF) was proposed. Firstly, the average value of hesitancy fuzzy elements was used to expand the data object with small hesitation. Secondly, the weights of data object before and after expansion were calculated by using the original information entropy and internal maximum difference, and the comprehensive attribute weight was determined according to the minimum discrimination information between the two weight vectors. Finally, with the goal of making the sum of weighted distances smaller, a center point construction method with constant hesitation was given. Experimental results on specific examples and synthetic datasets show that compared with the classic Hesitant Fuzzy Hierarchical Clustering algorithm (HFHC) and the recent Fuzzy Hierarchical Clustering Algorithm (FHCA), the proposed AHCHF increases the mean Silhouette Coefficient (SC) by 23.99% and 9.28% respectively, and shortens the running time by 27.18% and 6.40% averagely and respectively, proving that the proposed algorithm can effectively solve the problems of information distortion and poor objectivity of attribute weights, and improve the clustering effect and performance well.

    High utility itemset mining algorithm based on Markov optimization
    Xincheng ZHONG, Chang LIU, Xiumei ZHAO
    2023, 43(12):  3764-3771.  DOI: 10.11772/j.issn.1001-9081.2022121844
    Asbtract ( )   HTML ( )   PDF (1476KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    To address the problems that the High Utility Itemset Mining (HUIM) algorithms based on tree and link table structures often consume search spaces of orders of magnitude, and the evolutionary type-based mining algorithms fail to fully consider the interactions between variables, an HUIM algorithm based on Markov Optimization (HUIM-MOA) was proposed. Firstly, a bitmap matrix for expressing database and expectation vector encoding were used to achieve fast scanning of the database and efficient computation of utility values, respectively. Then, the Markov Network (MN) structure was estimated by computing the mutual information among dominant individuals and new populations were generated by using Gibbs sampling according to their local characteristics. Finally, population diversity preservation strategy and elite strategy were used to prevent the algorithm from falling into local optimum too quickly and to reduce the missing of high utility itemsets, respectively. Experimental results on real datasets show that compared with Bio-inspired High Utility Itemset Framework based on Particle Swarm Optimization (Bio-HUIF-PSO) algorithm, HUIM-MOA can find all the High Utility Itemsets (HUIs) when given a larger minimum threshold, with on average 12.5% improvement in convergence speed, 2.85 percentage point improvement in mined HUI number, and 14.6% reduction in running time. It can be seen that HUIM-MOA has stronger search performance than the evolutionary HUIM algorithm, which can effectively reduce the search time and improve the search quality.

    Missing value attention clustering algorithm based on latent factor model in subspace
    Xiaofei WANG, Shengli BAO, Jionghuan CHEN
    2023, 43(12):  3772-3778.  DOI: 10.11772/j.issn.1001-9081.2022121838
    Asbtract ( )   HTML ( )   PDF (1364KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    To solve the problems that traditional clustering algorithms are difficult to measure the sample similarity and have poor quality of filled data in the process of filling missing samples, a missing value attention clustering algorithm based on Latent Factor Model (LFM) in subspace was proposed. First, LFM was used to map the original data space to a low dimensional subspace to reduce the sparsity of samples. Then, the attention weight graph between different features was constructed by decomposing the feature matrix obtained from the original space, and the similarity calculation method between subspace samples was optimized to make the calculation of sample similarity more accurate and more generalized. Finally, to reduce the high time complexity in the process of sample similarity calculation, a multi-pointer attention weight graph was designed for optimization. The algorithm was tested on four proportional random missing datasets. On the Hand-digits dataset, compared with the KISC (K-nearest neighbors Interpolation Subspace Clustering) algorithm for high-dimensional feature missing data, when the missing data was 10%, the Accuracy (ACC) of the proposed algorithm was improved by 2.33 percentage points and the Normalized Mutual Information (NMI) was improved by 2.77 percentage points; when the missing data was 20%, the ACC of the proposed algorithm was improved by 0.39 percentage points, and the NMI was improved by 1.33 percentage points, which verified the effectiveness of the proposed algorithm.

    Multilabel feature selection algorithm based on Fisher score and fuzzy neighborhood entropy
    Lin SUN, Tianjiao MA, Zhan’ao XUE
    2023, 43(12):  3779-3789.  DOI: 10.11772/j.issn.1001-9081.2022121841
    Asbtract ( )   HTML ( )   PDF (1222KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    For that Fisher score model does not fully consider feature-label and label-label relations, and some neighborhood rough set models easily neglect the uncertainty of knowledge granulations in the boundary region, resulting in the low classification performance of these algorithms, a MultiLabel feature selection algorithm based on Fisher Score and Fuzzy neighborhood entropy (MLFSF) was proposed. Firstly, by using the Maximum Information Coefficient (MIC) to evaluate the feature-label association degree, the relationship matrix between features and labels was constructed, and the correlation between labels was analyzed by the relationship matrix of labels based on the adjusted cosine similarity. Secondly, a second-order strategy was given to obtain multiple second-order label relationship groups to reclassify the multilabel domain, where the strong correlation between labels was enhanced and the weak correlation between labels was weakened to obtain the score of each feature. The Fisher score model was improved to preprocess the multilabel data. Thirdly, the multilabel classification margin was introduced to define the adaptive neighborhood radius and neighborhood class, and the upper and lower approximation sets were constructed. On this basis, the multilabel rough membership degree function was presented, and the multilabel neighborhood rough set was mapped to the fuzzy set. Based on the multilabel fuzzy neighborhood, the upper and lower approximation sets and the multilabel fuzzy neighborhood rough set model were developed. Thus, the fuzzy neighborhood entropy and the multilabel fuzzy neighborhood entropy were defined to effectively measure the uncertainty of the boundary region. Finally, the Multilabel Fisher Score-based feature selection algorithm with second-order Label Correlation (MFSLC) was designed, and then the MLFSF was constructed. The experimental results applied to 11 multilabel datasets with the Multi-Label K-Nearest Neighbor (MLKNN) classifier show that when compared with six state-of-the-art algorithms including the Multilabel Feature Selection algorithm based on improved ReliefF (MFSR), MLFSF improves the mean of Average Precision (AP) by 2.47 to 6.66 percentage points; meanwhile, MLFSF obtains optimal values for all five evaluation metrics on most datasets.

    Cyber security
    Poisoning attack detection scheme based on generative adversarial network for federated learning
    Qian CHEN, Zheng CHAI, Zilong WANG, Jiawei CHEN
    2023, 43(12):  3790-3798.  DOI: 10.11772/j.issn.1001-9081.2022121831
    Asbtract ( )   HTML ( )   PDF (2367KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Federated Learning (FL) emerges as a novel privacy-preserving Machine Learning (ML) paradigm. However, the distributed training structure of FL is more vulnerable to poisoning attack, where adversaries contaminate the global model through uploading poisoning models, resulting in the convergence deceleration and the prediction accuracy degradation of the global model. To solve the above problem, a poisoning attack detection scheme based on Generative Adversarial Network (GAN) was proposed. Firstly, the benign local models were fed into the GAN to output testing samples. Then, the testing samples were used to detect the local models uploaded by the clients. Finally, the poisoning models were eliminated according to the testing metrics. Meanwhile, two test metrics named F1 score loss and accuracy loss were defined to detect the poisoning models and extend the detection scope from one single type of poisoning attacks to all types of poisoning attacks. Besides, a threshold determination method was designed to deal with misjudgment, so that the robust of misjudgment was confirmed. Experimental results on MNIST and Fashion-MNIST datasets show that the proposed scheme can generate high-quality testing samples, and then detect and eliminate poisoning models. Compared with the global models trained with the detection scheme based on directly gathering test data from clients and the detection scheme based on generating test data and using test accuracy as the test metric, the global model trained with the proposed scheme has significant accuracy improvement from 2.7 to 12.2 percentage points.

    Power battery safety warning based on time series anomaly detection
    Anqin ZHANG, Xiaohui WANG
    2023, 43(12):  3799-3805.  DOI: 10.11772/j.issn.1001-9081.2022111796
    Asbtract ( )   HTML ( )   PDF (2077KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Abnormal situations inside the vehicle battery cannot be predicted and warned in time, which leads to electric vehicle accidents and brings serious threats to drivers and passengers’ life and property safety. Aiming at the above problem, a Contrastive Transformer Encoder Decoder (CT-ED) model was proposed for multivariate time series anomaly detection. Firstly, different views of an instance were constructed through data augmentation, and the local invariant features of the data were captured by contrastive learning. Then, based on Transformer, the data were encoded from two perspectives of time dependence and feature dependence. Finally, the data were reconstructed by the decoder, and the reconstruction error was calculated as the anomaly score to detect anomalies of the machine under the actual operating conditions. Experimental results on SWaT, SMAP, MSL three public datasets and Electric Vehicle power battery (EV) dataset show that compared to the suboptimal model, the F1-scores of the proposed model increase by 6.5%, 1.8%, 0.9%, and 7.1% respectively.The above results prove that CT-ED is suitable for anomaly detection under different operating conditions, and balancing the precision and recall of anomaly detection.

    Advanced computing
    Multi-objective optimization model for unmanned aerial vehicles trajectory based on decomposition and trajectory search
    Junyan LIU, Feibo JIANG, Yubo PENG, Li DONG
    2023, 43(12):  3806-3815.  DOI: 10.11772/j.issn.1001-9081.2022121882
    Asbtract ( )   HTML ( )   PDF (1873KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The traditional Deep Learning (DL)-based multi-objective solvers have the problems of low model utilization and being easy to fall into the local optimum. Aiming at these problems, a Multi-objective Optimization model for Unmanned aerial vehicles Trajectory based on Decomposition and Trajectory search (DTMO-UT) was proposed. The proposed model consists of the encoding and decoding parts. First, a Device encoder (Dencoder) and a Weight encoder (Wencoder) were contained in the encoding part, which were used to extract the state information of the Internet of Things (IoT) devices and the features of the weight vectors. And the scalar optimization sub-problems that were decomposed from the Multi-objective Optimization Problem (MOP) were represented by the weight vectors. Hence, the MOP was able to be solved by solving all the sub-problems. The Wencoder was able to encode all sub-problems, which improved the utilization of the model. Then, the decoding part containing the Trajectory decoder (Tdecoder) was used to decode the encoding features to generate the Pareto optimal solutions. Finally, to alleviate the phenomenon of greedy strategy falling into the local optimum, the trajectory search technology was added in trajectory decoder, that was generating multiple candidate trajectories and selecting the one with the best scalar value as the Pareto optimal solution. In this way, the exploration ability of the trajectory decoder was enhanced during trajectory planning, and a better-quality Pareto set was found. The results of simulation experiments show that compared with the mainstream DL MOP solvers, under the condition of 98.93% model parameter quantities decreasing, the proposed model reduces the distribution of MOP solutions by 0.076%, improves the ductility of the solutions by 0.014% and increases the overall performance by 1.23%, showing strong ability of practical trajectory planning of DTMO-UT model.

    Route planning method of UAV swarm based on dynamic cluster particle swarm optimization
    Longbao WANG, Yinqi LUAN, Liang XU, Xin ZENG, Shuai ZHANG, Shufang XU
    2023, 43(12):  3816-3823.  DOI: 10.11772/j.issn.1001-9081.2022111763
    Asbtract ( )   HTML ( )   PDF (2693KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Route planning is very important for the task execution of Unmanned Aerial Vehicle (UAV) swarm, and the computation is usually complex in high dimensional scenarios. Swarm intelligence has provided a good solution for this problem. Particle Swarm Optimization (PSO) algorithm is especially suitable for route planning problem because of its advantages such as few parameters, fast convergence and simple operation. However, PSO algorithm has poor global search ability and is easy to fall into local optimum when applied to route planning. In order to solve the problems above and improve the effect of UAV swarm route planning, a Dynamic Cluster Particle Swarm Optimization (DCPSO) algorithm was proposed. Firstly, artificial potential field method and receding horizon control principle were used to model the task scenario of route planning problem of UAV swarm. Secondly, Tent chaotic map and dynamic cluster mechanism were introduced to further improve the global search ability and search accuracy. Finally, DCPSO algorithm was used to optimize the objective function of the model to obtain each trajectory point selection of UAV swarm. On 10 benchmark functions with different combinations of unimodal/multimodal and low-dimension/high-dimension, simulation experiments were carried out. The results show that compared with PSO algorithm, Pigeon-Inspired Optimization (PIO), Sparrow Search Algorithm (SSA) and Chaotic Disturbance Pigeon-Inspired Optimization (CDPIO) algorithm, DCPSO algorithm has better optimal value, mean value and variance, better search accuracy and stronger stability. Besides, the performance and effect of DCPSO algorithm were demonstrated in the route planning application instances of UAV swarm simulation experiments.

    Multi-objective optimization of minicells in distributed factories
    Chunfeng LIU, Zheng LI, Jufeng WANG
    2023, 43(12):  3824-3832.  DOI: 10.11772/j.issn.1001-9081.2022111772
    Asbtract ( )   HTML ( )   PDF (1527KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Due to differences in resource endowments and industrial policies among different regions, the role of distributed production in improving the competitiveness of manufacturing enterprises is very important. How to use distributed production to enhance the flexibility of mass customization is an important problem to be solved to boost consumer confidence. Combined with the idea of minicells — small manufacturing cells, in the distributed mixed production scenario with the multi-market and multi-product characteristics, an integrated model of distributed factory construction and production scheduling was proposed with the objectives to minimize the operating costs (e.g., labor and transportation costs) and minimize the makespan. By the proposed model, the minicell construction, worker and machine configuration, as well as production strategies for each batch of products were able to be solved. With the help of the proposed model, the enterprises were able to realize the quick release of production capacity and reasonable mixed flow production, so as to realize distributed manufacturing and sales that meet the multi-region, multi-product, and differentiated needs, and reduce the operating cost in the manufacturing process while guaranteeing the throughput. In addition, a Multi-Objective Particle Swarm Optimization (MOPSO) algorithm was designed to solve the proposed model, and was compared with Non-Dominated Sorting Genetic Algorithm Ⅱ (NSGA-Ⅱ) and Multi-Objective Simulated Annealing (MOSA) algorithm. The results of extensive numerical experiments show that MOPSO algorithm outperforms NSGA-Ⅱ and MOSA algorithm with the same running time in terms of three metrics: C-Metric (CM), Mean Ideal Distance (MID) and Maximum Spread (MS). The proposed algorithm can provide a high-quality decision-making scheme of production operation for the miniaturized distributed production system.

    Multi-robot task allocation algorithm combining genetic algorithm and rolling scheduling
    Fuqin DENG, Huanzhao HUANG, Chaoen TAN, Lanhui FU, Jianmin ZHANG, Tinlun LAM
    2023, 43(12):  3833-3839.  DOI: 10.11772/j.issn.1001-9081.2022121916
    Asbtract ( )   HTML ( )   PDF (2617KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The purpose of research on Multi-Robot Task Allocation (MRTA) is to improve the task completion efficiency of robots in smart factories. Aiming at the deficiency of the existing algorithms in dealing with large-scale multi-constrained MRTA, an MRTA Algorithm Combining Genetic Algorithm and Rolling Scheduling (ACGARS) was proposed. Firstly, the coding method based on Directed Acyclic Graph (DAG) was adopted in genetic algorithm to efficiently deal with the priority constraints among tasks. Then, the prior knowledge was added to the initial population of genetic algorithm to improve the search efficiency of the algorithm. Finally, a rolling scheduling strategy based on task groups was designed to reduce the scale of the problem to be solved, thereby solving large-scale problems efficiently. Experimental results on large-scale problem instances show that compared with the schemes generated by Constructive Heuristic Algorithm (CHA), MinInterfere Algorithm (MIA), and Genetic Algorithm with Penalty Strategy (GAPS), the scheme generated by the proposed algorithm has the average order completion time shortened by 30.02%, 16.86% and 75.65% respectively when the number of task groups is 20, which verifies that the proposed algorithm can effectively shorten the average waiting time of orders and improve the efficiency of multi-robot task allocation.

    Solving robot path planning problem by adaptively adjusted Harris hawk optimization algorithm
    Lin HUANG, Qiang FU, Nan TONG
    2023, 43(12):  3840-3847.  DOI: 10.11772/j.issn.1001-9081.2022121847
    Asbtract ( )   HTML ( )   PDF (1437KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the problem that the heuristic algorithms have unstable path lengths and are easy to fall into local minimum in the process of robot path planning, an Adaptively Adjusted Harris Hawk Optimization (AAHHO) algorithm was proposed. Firstly, the convergence factor adjustment strategy was used to adjust the balance between the global search stage and the local search stage, and the natural constant was used as the base to improve the search efficiency and convergence accuracy. Then, in the global search phase, the elite cooperation guided search strategy was adopted, by three elite Harris hawks cooperatively guiding other individuals to update the positions, so that the search performance was enhanced, and the information exchange among the populations was enhanced through the three optimal positions. Finally, by simulating the intraspecific competition strategy, the ability of the Harris hawks to jump out of the local optimum was improved. The comparative experimental results of function testing and robot path planning show that the proposed algorithm is superior to comparison algorithms such as IHHO(Improve Harris Hawk Optimization) and CHHO(Chaotic Harris Hawk Optimization), in both function testing and path planning, and it has better effectiveness, feasibility and stability in robot path planning.

    Integrated scheduling considering automated guided vehicle charging strategy based on improved NSGA-Ⅱ
    Hairong XUE, Xiaolong HAN
    2023, 43(12):  3848-3855.  DOI: 10.11772/j.issn.1001-9081.2022121923
    Asbtract ( )   HTML ( )   PDF (3405KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the power problem of Automated Guided Vehicle (AGV) in the process of performing tasks in Automated Container Terminal (ACT), an integrated scheduling considering AGV charging strategy based on improved Non-dominated Sorting Genetic Algorithm-Ⅱ (NSGA-Ⅱ) was proposed. Firstly, considering the power consumption of AGV under different operating statuses in the integrated scheduling mode of quay crane, yard crane and AGV, a multi-objective mixed programming model with the goal of minimizing the completion time and total power consumption was established. Secondly, to improve the performance of the traditional NSGA-Ⅱ, an adaptive NSGA-Ⅱ was designed and compared with CPLEX solver and Multi-Objective Partical Swarm Optimization (MOPSO) algorithm on performance. Finally, different charging strategies and equipment number ratios of AGV were designed for experimental research. The experimental results of algorithm comparison show that the solution results of the adaptive NSGA-Ⅱ are improved by 2. 80% and 2. 63% respectively on the two objectives proposed compared with NSGA-Ⅱ. The experimental results of applying the adaptive NSGA-Ⅱ to study the ratio of charging strategies and equipment number ratios show that increasing AGV charging number can reduce AGV charging time, and adjusting the ratio of the equipment number to 3:3:9 and 3:7:3 lead to the highest time utilization of yard crane and AGV respectively. It can be seen that the AGV charging strategy and equipment number ratio can influence the terminal integrated scheduling with multiple equipment.

    Band sparse matrix multiplication and efficient GPU implementation
    Li LIU, Changbo CHEN
    2023, 43(12):  3856-3867.  DOI: 10.11772/j.issn.1001-9081.2022111720
    Asbtract ( )   HTML ( )   PDF (3222KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Sparse-dense Matrix Multiplication (SpMM) is widely used in the fields such as scientific computing and deep learning, and it is of great importance to improve its efficiency. For a class of sparse matrices with band feature, a new storage format BRCV (Banded Row Column Value) and an SpMM algorithm based on this format as well as an efficient Graphics Processing Unit (GPU) implementation were proposed. Due to the fact that each sparse band can contain multiple sparse blocks, the proposed format can be seen as a generalization of the block sparse matrix format. Compared with the commonly used CSR (Compressed Sparse Row) format, BRCV format was able to significantly reduce the storage complexity by avoiding redundant storage of column indices in sparse bands. At the same time, the GPU implementation of SpMM based on BRCV format was able to make more efficient use of GPU’s shared memory and improve the computational efficiency of SpMM algorithm by reusing the rows of both sparse and dense matrices. For randomly generated band sparse matrices, experimental results on two different GPU platforms show that BRCV outperforms not only cuBLAS (CUDA Basic Linear Algebra Subroutines), but also cuSPARSE based on CSR and block sparse formats. Specifically, compared with cuSPARSE based on CSR format, BRCV has the maximum speedup ratio of 6.20 and 4.77 respectively. Moreover, the new implementation was applied to accelerate the SpMM operator in Graph Neural Network (GNN). Experimental results on real application datasets show that BRCV outperforms cuBLAS and cuSPARSE based on CSR format, also outperforms cuSPARSE based on block sparse format in most cases. In specific, compared with cuSPARSE based on CSR format, BRCV has the maximum speedup ratio reached 4.47. The above results indicate that BRCV can improve the efficiency of SpMM effectively.

    Improved TLBO algorithm with adaptive competitive learning
    Peichong WANG, Haojing FENG, Lirong LI
    2023, 43(12):  3868-3874.  DOI: 10.11772/j.issn.1001-9081.2023010025
    Asbtract ( )   HTML ( )   PDF (1194KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    For that the Teaching-Learning-Based Optimization (TLBO) algorithm has some problems, such as prematurity and poor solution accuracy, in solving high-dimensional optimization problems, an Improved TLBO algorithm with Adaptive Competitive learning (ITLBOAC) was proposed. Firstly, a weighted parameter with nonlinear change was introduced into the “teaching” operator to determine the ability of the current individual to maintain its own state and adjust the attitude of the current individual towards learning from teachers. As a result, the current individual learnt more from the teacher in the early stage to improve its own state quickly, and kept the state of itself more in the later stage to slow down the influence of the teacher on it. Then, based on ecological cooperation and competition mechanisms, a “learning” operator based on adaptive competition between nearest neighbor individuals was introduced. To make the current individual chose its near neighbors and the individuals eventually shifted from cooperative evolution to competitive learning. Test results on 12 Benchmark test functions show that compared with four improved TLBO algorithms, the proposed algorithm is better in terms of accuracy of solutions, stability and convergence speed, and is much better than TLBO algorithm at the same time, which verify that the proposed algorithm is suitable for solving high-dimensional continuous optimization problems. Test results with compression spring and three-bar truss design problems selected to test show that the optimal values obtained by ITLBOAC decreased by 3.03% and 0.34% respectively, compared with those obtained by TLBO algorithm. It can be seen that ITLBOAC is a trustworthy algorithm in solving constrained engineering optimization problems.

    Approximate evaluation method of k-ary (n-1)-cube subnetwork reliability
    Kai FENG, Jiande LI, Zhangjian JI
    2023, 43(12):  3875-3881.  DOI: 10.11772/j.issn.1001-9081.2022111719
    Asbtract ( )   HTML ( )   PDF (2399KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The implementation of the functions of a multiprocessor system relies heavily on the topological properties of the interconnection network of this system. The subnetwork reliability of k-ary n-cube network is an important factor that needs to be taken into account when the computing tasks are processed by the multiprocessor systems constructed with k-ary n-cube as topological structure. In order to accurately and efficiently measure the reliability of the k-ary (n-1)-cube subnetwork in a k-ary n-cube under the probabilistic fault condition, an approximate method to evaluate the reliability of k-ary (n-1)-cube subnetwork based on the Back Propagation (BP) neural network was proposed. Firstly, the generation method for dataset to train BP neural network was given by the aid of the Monte Carlo simulation method and the known upper and lower bounds on the reliability of the k-ary (n-1)-cube subnetwork. Then, the BP neural network model for evaluating the reliability of the k-ary (n-1)-cube subnetwork was constructed on the basis of the generated training dataset. Finally, the approximate evaluation results of the k-ary (n-1)-cube subnetwork reliability obtained by the BP neural network model were analyzed and compared with the results obtained by the approximate calculation formula and the evaluation method based on Monte Carlo simulation. The results obtained by the proposed method were more accurate compared with the approximate calculation formula, and the evaluation time of the proposed method was reduced by about 59% on average compared with the evaluation method based on Monte Carlo simulation. Experimental results show that the proposed method has certain advantages in balancing accuracy and efficiency.

    Network and communications
    Routing lookup algorithm with variable-length address based on AVL tree and Bloom filter
    Yongjin HUANG, Yifang QIN, Xu ZHOU, Xinqing ZHANG
    2023, 43(12):  3882-3889.  DOI: 10.11772/j.issn.1001-9081.2022121915
    Asbtract ( )   HTML ( )   PDF (2064KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The variable-length address is one of the important research content in the field of future network. Aiming at the low efficiency of traditional routing lookup algorithms for variable-length address, an efficient routing lookup algorithm suitable for variable-length addresses based on balanced binary tree — AVL (Adelson-Velskii and Landis) tree and Bloom filter, namely AVL-Bloom algorithm, was proposed. Firstly, multiple off-chip hash tables were used to separately store route entries with the same number of prefix bits and their next-hop information in view of the flexible and unbounded characteristics of the variable-length address. Meanwhile, the on-chip Bloom filter was utilized for speeding up the search for route prefixes that were likely to match. Secondly, in order to solve the problem that the routing lookup algorithms based on hash technology need multiple hash comparisons when searching for the route with the longest prefix, the AVL tree technology was introduced, that was, the Bloom filter and hash table of each group of route prefix set were organized through AVL tree, so as to optimize the query order of route prefix length and reduce the number of hash calculations and then decrease the search time. Finally, comparative experiments of the proposed algorithm with the traditional routing lookup algorithms such as METrie (Multi-Entrance-Trie) and COBF (Controlled prefix and One-hashing Bloom Filter) were conducted on three different variable-length address datasets. Experimental results show that the search speed of AVL-Bloom algorithm is significantly faster than those of METrie and COBF algorithms, and the query time is reduced by nearly 83% and 64% respectively. At the same time, AVL-Bloom algorithm can maintain stable search performance under the condition of large change in routing table entries, and is suitable for routing lookup and forwarding with variable-length addresses.

    Low-complexity generalized space shift keying signal detection algorithm based on compressed sensing
    Xinhe ZHANG, Haoran TAN, Wenbo LYU
    2023, 43(12):  3890-3895.  DOI: 10.11772/j.issn.1001-9081.2022121808
    Asbtract ( )   HTML ( )   PDF (1143KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    As a simplified version of Spatial Modulation (SM), Generalized Space Shift Keying (GSSK) has been widely used in massive Multiple-Input Multiple-Output (MIMO) systems. It can better solve the problems such as Inter-Channel Interference (ICI), Inter-Antenna Synchronization (IAS), and multiple Radio Frequency (RF) links in traditional MIMO technology. To solve the problem of high computational complexity of the Maximum Likelihood (ML) detection algorithm for GSSK systems, a low-complexity GSSK signal detection algorithm based on Compressed Sensing (CS) theory was proposed by combining Subspace Tracking (SP) and ML detection algorithms in CS, and presetting the threshold. First, the improved SP algorithm was used to obtain partial Transmit Antenna Combinations (TACs). Secondly, the set of search antennas was shrunk by deleting partial antenna combinations. Finally, the ML algorithm and the preset threshold were used to estimate the TACs. The results of simulation experiments show that the computational complexity of the proposed algorithm is significantly lower than that of ML detection algorithm, and the Bit Error Rate (BER) performance is almost the same as that of ML detection algorithm, which verify the effectiveness of the proposed algorithm.

    Computer software technology
    Code search model based on collaborative fusion network
    Qihong SONG, Jianxun LIU, Haize HU, Xiangping ZHANG
    2023, 43(12):  3896-3902.  DOI: 10.11772/j.issn.1001-9081.2022111783
    Asbtract ( )   HTML ( )   PDF (1457KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Searching and reusing relevant code can significantly improve software development efficiency. The deep learning-based code search models usually embed code pieces and query statements into the same vector space and then match and output the relevant code by computing cosine similarity; however, most of these models ignore the collaborative information between code pieces and query statements. To fully represent semantic information, a collaborative fusion-based code search model named BofeCS was proposed. Firstly, BERT (Bidirectional Encoder Representations from Transformers) model was utilized to extract the semantic information of the input sequences and then represent it as vectors. Secondly, a collaborative fusion network was constructed to extract the token-level collaborative information between code pieces and query statements. Finally, a residual network was built to alleviate the semantic information loss during the representation process. The multi-lingual dataset CodeSearchNet was used to carry out experiments to evaluate the effectiveness of BofeCS. Experimental results show that BofeCS can significantly improve the accuracy of code search and outperform the baseline models, UNIF (embedding UNIFication), TabCS (Two-stage Attention-Based model for Code Search), and MRCS (Multimodal Representation for neural Code Search), in Mean Reciprocal Rank (MRR), Normalized Discounted Cumulative Gain (NDCG), and Top k Success hit Rate (SR@k), where the MRR values are improved by 95.94%, 52.32%, and 16.95%, respectively.

    Multimedia computing and computer simulation
    Single image super-resolution method based on residual shrinkage network in real complex scenes
    Ying LI, Chao HUANG, Chengdong SUN, Yong XU
    2023, 43(12):  3903-3910.  DOI: 10.11772/j.issn.1001-9081.2022111697
    Asbtract ( )   HTML ( )   PDF (3309KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    There are very few paired high and low resolution images in the real world. The traditional single image Super-Resolution (SR) methods typically use pairs of high-resolution and low-resolution images to train models, but these methods use the way of synthetizing dataset to obtain training set, which only consider bilinear downsampling as image degradation process. However, the image degradation process in the real word is complex and diverse, and traditional image super-resolution methods have poor reconstruction performance when facing real unknown degraded images. Aiming at those problems, a single image super-resolution method was proposed for real complex scenes. Firstly, high- and low-resolution images were captured by the camera with different focal lengths, and these images were registered as image pairs to form a dataset CSR(Camera Super-Resolution dataset) of various scenes. Secondly, to simulate the image degradation process in the real world as much as possible, the image degradation model was improved by the parameter randomization of degradation factors and the nonlinear combination degradation. Besides, the dataset of high- and low-resolution image pairs and the image degradation model were combined to synthetize training set. Finally, as the degradation factors were considered in the dataset, residual shrinkage network and U-Net were embedded into the benchmark model to reduce the redundant information caused by degradation factors in the feature space as much as possible. Experimental results indicate that compared with the BSRGAN (Blind Super-Resolution Generative Adversarial Network) method, under complex degradation conditions, the proposed method improves the PSNR by 0.7 dB and 0.14 dB, and improves SSIM by 0.001 and 0.031 respectively on the RealSR and CSR test sets. The proposed method has better objective indicators and visual effect than the existing methods on complex degradation datasets.

    Outdoor scene point cloud segmentation model based on graph model and attention mechanism
    Feiyu LIAN, Liang ZHANG, Jiedong WANG, Yukang JIN, Yu CHAI
    2023, 43(12):  3911-3917.  DOI: 10.11772/j.issn.1001-9081.2022111704
    Asbtract ( )   HTML ( )   PDF (2141KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the problem that it is difficult to distinguish similar land types in outdoor scenes with multiple objects and complex spatial topological relationships, an A-Edge-SPG (Attention-EdgeConv SuperPoint Graph) graph neural network combining graph model and attention mechanism module was proposed. Firstly, the superpoints were segmented by the combination of graph cut and geometric features. Secondly, the local adjacency graph was constructed inside the superpoint to capture the context information of the point cloud in the scene and use the attention mechanism module to highlight the key information. Finally, a SuperPoint Graph (SPG) model was constructed, and the features of hyperpoints and hyperedges were aggregated by Gated Recurrent Unit (GRU) to realize accurate segmentation among different land types of point cloud. On Semantic3D dataset, the semantic segmentation effect of A-Edge-SPG model and SPG-Net (SPG neural Network) model was compared and analyzed. Experimental results show that compared with the SPG model, A-Edge-SPG model improves the Overall segmentation Accuracy(OA), mean Intersection over Union (mIoU) and mean Average Accuracy (mAA) by 1.8, 5.1 and 2.8 percentage points respectively, and significantly improves the segmentation accuracy of similar land types such as high vegetation and dwarf vegetation, improving the effect of distinguishing similar land types.

    Rectal cancer segmentation network based on adjacent slice attention fusion
    Donglei LAN, Xiaodong WANG, Yu YAO, Xin WANG, Jitao ZHOU
    2023, 43(12):  3918-3926.  DOI: 10.11772/j.issn.1001-9081.2023010045
    Asbtract ( )   HTML ( )   PDF (2681KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the problem that the target regions of rectal cancer show different sizes, shapes, textures, and boundary clarity on Magnetic Resonance Imaging (MRI) images, to overcome the individual variability among patients and improve the segmentation accuracy, an Adjacent Slice Attention Fusion Network for rectal cancer segmentation (ASAF-Net) was proposed. Firstly, using High Resolution Network (HRNet) as the backbone network, the high-resolution feature representation was maintained during the feature extraction process, thereby reducing the loss of semantic information and spatial location information. Secondly, the multi-scale contextual semantic information between adjacent slices was fused and enhanced by the Adjacent Slice Attention Fusion (ASAF) module, so that the network was able to learn the spatial features between adjacent slices. Finally, in the decoder, the co-training of Fully Convolutional Network (FCN) and Atrous Spatial Pyramid Pooling (ASPP) segmentation heads was carried out, and the large differences between adjacent slices during training was reduced by adding consistency constraints between adjacent slices as an auxiliary loss. Experimental results show that compared with HRNet, ASAF-Net improves the mean Intersection over Union (IoU) and mean Dice Similarity Coefficient (DSC) by 1.68 and 1.26 percentage points, respectively, and reduces the 95% mean Hausdorff Distance (HD) by 0.91 mm. At the same time, ASAF-Net can achieve better segmentation results in both internal filling and edge prediction of multi-objective target regions in rectal cancer MRI image, and helps to improve physician efficiency in clinical auxiliary diagnosis.

    Book spine segmentation algorithm based on improved DeepLabv3+ network
    Xiaofei JI, Kexin ZHANG, Lirong TANG
    2023, 43(12):  3927-3932.  DOI: 10.11772/j.issn.1001-9081.2022121887
    Asbtract ( )   HTML ( )   PDF (2364KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The location of books is one of the critical technologies to realize the intelligent development of libraries, and the accurate book spine segmentation algorithm has become a major challenge to achieve this goal. Based on the above solution, an improved book spine segmentation algorithm based on improved DeepLabv3+ network was proposed, aiming to solve the difficulties in book spine segmentation caused by dense arrangement, skew angles of books, and extremely similar book spine textures. Firstly, to extract more dense pyramid features of book images, the Atrous Spatial Pyramid Pooling (ASPP) in the original DeepLabv3+ network was replaced by the multi-dilation rate and multi-scale DenseASPP (Dense Atrous Spatial Pyramid Pooling) module. Secondly, to solve the problem of insensitivity of the original DeepLabv3+ network to the segmentation boundaries of objects with large aspect ratios, Strip Pooling (SP) module was added to the branch of the DenseASPP module to enhance the strip features of book spines. Finally, based on the Multi-Head Self-Attention (MHSA) mechanism in ViT (Vision Transformer), a global information enhancement-based self-attention mechanism was proposed to enhance the network’s ability to obtain long-distance features. The proposed algorithm was tested and compared on an open-source database, and the experimental results show that compared with the original DeepLabv3+ network segmentation algorithm, the proposed algorithm improves the Mean Intersection over Union (MIoU) by 1.8 percentage points on the nearly vertical book spine database and by 4.1 percentage points on the skewed book spine database, and the latter MIoU of the proposed algorithm achieves 93.3%. The above confirms that the proposed algorithm achieves accurate segmentation of book spine targets with certain skew angles, dense arrangement, and large aspect ratios.

    Parking space detection method based on self-supervised learning HOG prediction auxiliary task
    Lei LIU, Peng WU, Kai XIE, Beizhi CHENG, Guanqun SHENG
    2023, 43(12):  3933-3940.  DOI: 10.11772/j.issn.1001-9081.2022111687
    Asbtract ( )   HTML ( )   PDF (2364KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    In the intelligent parking space management system, a decrease in accuracy and effectiveness of parking space prediction can be caused by factors such as illumination changes and parking space occlusion. To overcome this problem, a parking space detection method based on self-supervised learning HOG (Histogram of Oriented Gradient) prediction auxiliary task was proposed. Firstly, a self-supervised learning auxiliary task to predict the HOG feature in occluded part of image was designed, the visual representation of the image was learned more fully and the feature extraction ability of the model was improved by using the MobileViTBlock (light-weight, general-purpose, and Mobile-friendly Vision Transformer Block) to synthesize the global information of the image. Then, an improvement was made to the SE (Squeeze-and-Excitation) attention mechanism, thereby enabling the model to achieve or even exceed the effect of the original SE attention mechanism at a lower computational cost. Finally, the feature extraction part trained by the auxiliary task was applied to the downstream classification task for parking space status prediction. Experiments were carried out on the mixed dataset of PKLot and CNRPark. The experimental results show that the proposed model has the accuracy reached 97.49% on the test set; compared to RepVGG, the accuracy of occlusion prediction improves by 5.46 percentage points, which represents a great improvement compared with other parking space detection algorithms.

    Frontier and comprehensive applications
    Anesthesia resuscitation object detection method based on improved single shot multibox detector
    Ronghao LUO, Zhiyou CHENG, Chuanjian WANG, Siqian LIU, Zhentian WANG
    2023, 43(12):  3941-3946.  DOI: 10.11772/j.issn.1001-9081.2022121917
    Asbtract ( )   HTML ( )   PDF (2727KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    The target detection model of anesthesia resuscitation is often used to help medical staff to perform resuscitation detection on anesthetized patients. The targets of facial actions during patient resuscitation are small and are not obvious, and the existing Single Shot multibox Detector (SSD) is difficult to accurately detect the facial micro-action features of patients in real time. Aiming at the problem that the original model has low detection speed and is easy to have missed detection, an anesthesia resuscitation object detection method based on improved SSD was proposed. Firstly, the backbone network VGG (Visual Geometry Group)16 of the original SSD was replaced by the lightweight backbone network MobileNetV2, and the standard convolutions were replaced by the depthwise separable convolutions. At the same time, the calculation method of first increasing and then reducing the dimension of the extracted features from patient photos was used to reduce computational cost, thereby improving detection speed of the model. Secondly, the Coordinate Attention (CA) mechanism was integrated into the feature layers with different scales extracted by the SSD, and the ability of the feature map to extract key information was improved by weighting the channel and location information, so that the network positioning and classification performance was optimized. Finally, comparative experiments were carried out on three datasets: CEW(Closed Eyes in the Wild), LFW(Labeled Faces in the Wild), and HAPF(Hospital Anesthesia Patient Facial). Experimental results show that the mean Average Precision (AP) of the proposed model reaches 95.23%, and the detection rate of photos is 24 frames per second, which are 1.39 percentage points higher and 140% higher than those of the original SSD model respectively. Therefore, the improved model has the effect of real-time accurate detection in anesthesia resuscitation detection, and can assist medical staff in resuscitation detection.

    Fingerprint positioning method based on measurement report signal clustering
    Haiyong ZHANG, Xianjin FANG, Enwan ZHANG, Baoyu LI, Chao PENG, Jianxiang MU
    2023, 43(12):  3947-3954.  DOI: 10.11772/j.issn.1001-9081.2023010005
    Asbtract ( )   HTML ( )   PDF (2357KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the problems of low positioning precision and efficiency of fingerprint positioning methods based on Weighted K-Nearest Neighbor (WKNN) and machine learning algorithms, a fingerprint positioning method based on Measurement Report (MR) signal clustering was proposed. Firstly, MR signals were divided into three attributes: indoor, road and outdoor. Then, by using the Geographic Information System (GIS) information, the grids were divided into building, road and outdoor sub-regions, and MR data with different attributes were placed in the sub-regions with corresponding attributes. Finally, with the help of K-Means clustering algorithm, MR signals in the grid were clustered and analyzed to create virtual sub-regions under the sub-region, and WKNN algorithm was used to match MR test samples. Besides, the average positioning accuracy was calculated by using the Euclidean distance, and the positioning performance of the proposed method was tested by some MR data in the production environment. Experimental results show that the proportion of 50 m positioning error of the proposed method is 71.21%, which is 2.64 percentage points higher than that of WKNN algorithm, and the average positioning error of the proposed method is 44.73 m, which is 7.60 m lower than that of WKNN algorithm. It can be seen that the proposed method has good positioning precision and efficiency, and can meet the positioning requirements of MR data in the production environment.

    Customs risk control method based on improved butterfly feedback neural network
    Zhenggang WANG, Zhong LIU, Jin JIN, Wei LIU
    2023, 43(12):  3955-3964.  DOI: 10.11772/j.issn.1001-9081.2022121873
    Asbtract ( )   HTML ( )   PDF (2964KB) ( )  
    Figures and Tables | References | Related Articles | Metrics

    Aiming at the problems of low efficiency, low accuracy, excessive occupancy of human resources and intelligent classification algorithm miniaturization deployment requirements in China Customs risk control methods at this stage, a customs risk control method based on an improved Butterfly Feedback neural Network Version 2 (BFNet-V2) was proposed. Firstly, the Filling in Code (FC) algorithm was used to realize the semantic replacement of the customs tabular data to the analog image. Then, the analog image data was trained by using the BFNet-V2. The regular neural network structure was composed of left and right links, different convolution kernels and blocks, and small block design, and the residual short path was added to improve the overfitting and gradient disappearance. Finally, a Historical momentum Adaptive moment estimation algorithm (H-Adam) was proposed to optimize the gradient descent process and achieve a better adaptive learning rate adjustment, and classify customs data. Xception (eXtreme inception), Mobile Network (MobileNet), Residual Network (ResNet), and Butterfly Feedback neural Network (BF-Net) were selected as the baseline network structures for comparison. The Receiver Operating Characteristic curve (ROC) and the Precision-Recall curve (PR) of the BFNet-V2 contain the curves of the baseline network structures. Taking Transfer Learning (TL) as an example, compared with the four baseline network structures, the classification accuracy of BFNet-V2 increases by 4.30%,4.34%,4.10% and 0.37% respectively. In the process of classifying real-label data, the misjudgment rate of BFNet-V2 reduces by 70.09%,57.98%,58.36% and 10.70%, respectively. The proposed method was compared with eight classification methods including shallow and deep learning methods, and the accuracies on three datasets increase by more than 1.33%. The proposed method can realize automatic classification of tabular data and improve the efficiency and accuracy of customs risk control.

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