模式识别与人工智能
Home      About Journal      Editorial Board      Instructions      Ethics Statement      Contact Us                   中文
Pattern Recognition and Artificial Intelligence
22 Judgement and Disposal of Academic Misconduct Article
22 Copyright Transfer Agreement
22 Proof of Confidentiality
22 Requirements for Electronic Version
More....
22 Chinese Association of Automation
22 National ResearchCenter for Intelligent Computing System
22 Institute of Intelligent Machines,Chinese Academy of Sciences
More....
 
 
2020 Vol.33 Issue.4, Published 2020-04-25
Papers and Reports    Researches and Applications   
   
Papers and Reports
287 Social Media Text Classification Method Based on Character-Word Feature Self-attention Learning
WANG Xiaoli, YE Dongyi
Long tail effect and excessive out-of-vocabulary(OOV) words in social media texts result in severe feature sparsity and reduce classification accuracy. To solve the problem, a social media text classification method based on character-word feature self-attention learning is proposed. Global features are constructed at the character level to learn attention weight distribution, and the existing multi-head attention mechanism is improved to reduce parameter scale and computational complexity. To further analyze character-word feature fusion, OOV sensitivity is proposed to measure the impact of OOV words on different types of features. Experiments on several social media text classification tasks indicate that the effectiveness and classification accuracy of the proposed method are obviously improved in terms of fusing word features and character features. Moreover, the quantitative results of OOV vocabulary sensitivity index verify the feasiblity and effectiveness of the proposed method.
2020 Vol. 33 (4): 287-294 [Abstract] ( 651 ) [HTML 1KB] [ PDF 654KB] ( 413 )
295 Personalized Recommendation Algorithm Based on Deep Neural Network and Weighted Implicit Feedback
XUE Feng, LIU Kai, WANG Dong, ZHANG Haobo

In singular value decomposition++(SVD++), inner product of user and item feature vector is regarded as user's rating of items. However, inner product cannot capture the high-order nonlinear relationship between the user and the item. In addition, the contribution of different interactive items cannot be distinguished when user's implicit feedback is incorporated in SVD++. A recommendation algorithm based on deep neural network and weighted implicit feedback is proposed to solve the two problems. Deep neural network is adopted to model the relationship between the user and the object and attention mechanism is utilized to calculate the weight of historical interactive items in modeling user's implicit feedback. Experiments on public datasets verify the effectiveness of the proposed algorithm.
2020 Vol. 33 (4): 295-302 [Abstract] ( 543 ) [HTML 1KB] [ PDF 818KB] ( 412 )
303 Joint Sparse Representation Fusing Hierarchical Deep Network of Hyperspectral Image Classification
WANG Junhao, YAN Deqin, LIU Deshan, YAN Huicong
In joint sparse representation of hyperspectral image classification, once the local window of each pixel includes samples from different categories, the dictionary atoms and testing samples are easily affected by samples from different categories with same spectrum and the classification performance is seriously decreased. According to the characteristics of hyperspectral image , an algorithm of joint sparse representation fusing hierarchical deep network is proposed. Discriminative spectral information and spatial information are extracted by alternating spectral and spatial feature learning operations, and then a dictionary with spatial spectral features is constructed for joint sparse representation. In the classification process, the correlation coefficient between the dictionary and the testing samples is combined with classification error to make decisions. Experiments on two hyperspectral remote sensing datasets verify the effectiveness of the proposed algorithm.
2020 Vol. 33 (4): 303-312 [Abstract] ( 412 ) [HTML 1KB] [ PDF 2140KB] ( 251 )
313 Manifold Spectral Clustering Image Segmentation Algorithm    Based on Local Geometry Features
ZHANG Rongguo, YAO Xiaoling, ZHAO Jian, HU Jing, LIU Xiaojun
To improve the accuracy and timeliness of spectral clustering image segmentation,an algorithm of manifold spectral clustering image segmentation based on local geometry features is proposed. Firstly, considering the manifold structure of image data, the relationship of data intrinsic dimensions is obtained by performing spectral clustering based on local principal components analysis in the k-nearest neighbor region of data points. Then, the local linear reconstruction technique in manifold learning is introduced, and the similarity of local tangent space between data is obtained via mixed linear analyzers, and the similarity matrix with local geometric features is constructed by merging the intrinsic dimension and the local tangent space. Nyström technique is utilized to approximate eigenvectors of the image to be segmented, and spectral clustering is performed on the constructed k principal eigenvectors. Finally, experiments on Berkeley dataset show the advantages of the proposed algorithm in accuracy and timeliness.
2020 Vol. 33 (4): 313-324 [Abstract] ( 401 ) [HTML 1KB] [ PDF 1401KB] ( 372 )
325 Face Super-Resolution Reconstruction Method Fusing Reference Image
FU Lihua, LU Zhongshan, SUN Xiaowei, ZHAO Yu, ZHANG Bo
While low-resolution face images are reconstructed via deep learning based super-resolution reconstruction method, some problems emerge, such as blurred reconstructed images and obvious difference between reconstructed images and real images. Aiming at these problems, a face super-resolution reconstruction method fusing reference image is proposed to reconstruct low-resolution human face images effectively. The multi-scale features of reference image are extracted by reference image feature extraction subnet to retain the detail information of key parts and remove the redundant information, such as facial contour and facial expression. Based on the multi-scale features of reference image, the step-by-step super-resolution main network fills the features to low-resolution face image step by step. Finally, the high-resolution face image is generated. Experiments on datasets indicate that the proposed method reconstructs low-resolution face images effectively with good robustness.
2020 Vol. 33 (4): 325-336 [Abstract] ( 441 ) [HTML 1KB] [ PDF 2006KB] ( 395 )
Researches and Applications
337 Dynamic Knowledge Graph Inference Based on Multiple Relational Cyclic Events
CHEN Hao, LI Yongqiang, FENG Yuanjing
The reasoning ability of most existing dynamic knowledge map reasoning methods under the same time and multiple relationships is limited . Aiming at this problem, a method of dynamic knowledge graph inference based on multi-relational cyclic events(Multi-Net) is proposed. The improved multi-relational proximity aggregator is employed to fuse target entity neighborhood information to obtain more accurate representation of entity neighborhood vector, and Multi-Net is simplified by optimizing information fusion, and the ability to handle the conflict of relations between two entities in a specific scope is improved by adding the relationship prediction task to Multi-Net. Experiments of entity prediction and relationship prediction on large real datasets indicate that Multi-Net improves the reasoning ability of dynamic knowledge maps effectively.
2020 Vol. 33 (4): 337-343 [Abstract] ( 465 ) [HTML 1KB] [ PDF 626KB] ( 586 )
344 Latent Low-Rank Sparse Multi-view Subspace Clustering
ZHANG Zhuohan, CAO Rongwei, LI Chen, CHENG Shiqing
To solve the problem of multi-view clustering, a latent low-rank sparse multi-view subspace clustering(LLSMSC) algorithm is proposed. A latent space shared by all views is constructed to explore the complementary information of multi-view data. The global and local structure of multi-view data can be captured to attain promising clustering results by imposing low-rank constraint and sparse constraint on the implicit latent subspace representation simultaneously. An algorithm based on augmented Lagrangian multiplier with alternating direction minimization strategy is employed to solve the optimization problem. Experiments on six benchmark datasets verify the effectiveness and superiority of LLSMSC.
2020 Vol. 33 (4): 344-352 [Abstract] ( 396 ) [HTML 1KB] [ PDF 831KB] ( 308 )
353 An Algorithm for Road Boundary Extraction and Obstacle Detection Based on 3D Lidar
WANG Can, KONG Bin, YANG Jing, WANG Zhiling, ZHU Hui
To extract relevant road information quickly and effectively for intelligent vehicles in various road environments, an algorithm for real-time road boundary extraction and obstacle detection based on three-dimensional (3D) lidar is proposed. Firstly, 3D lidar point cloud data is rasterized and filtered, and the single beam laser point cloud spatial segmentation method is employed for spatial analysis to obtain the point cloud smoothness characteristic image. Then, the adaptive direction search algorithm is adopted to obtain the road boundary feature points and perform cluster analysis and curve fitting. Finally, the point cloud in the passable area is clustered and segmented under the road boundary constraint to obtain the obstacle position information. Experiments show that the proposed algorithm extracts road boundary and obstacle location information accurately in real time, and it meets the requirements of environment modeling and path planning for intelligent vehicle.
2020 Vol. 33 (4): 353-362 [Abstract] ( 718 ) [HTML 1KB] [ PDF 3109KB] ( 989 )
363 Generative High-Resolution Image Inpainting with Parallel Adversarial Network and Multi-condition Fusion
SHAO Hang, WANG Yongxiong
Regions with artifacts and semantic inaccuracy are often caused by existing image inpainting algorithms. Moreover, the inpainting effect is limited for images with large missing regions and high-resolution. Therefore, a two-stage image inpainting approach based on parallel adversarial network and multi-condition fusion is proposed in this paper. Firstly, an improved deep residual network is utilized to fill the corrupted image. The first-stage adversarial network is employed to complete the image edge map. Next, the color feature of the filled image is extracted and fused with the completed edge image. Then, the fused image is applied as the condition label of the second-stage adversarial network. Finally, the inpainting result is obtained by the second-stage network with a contextual attention module. Experiments on multiple public datasets demonstrate that realistic inpainting results can be obtained by the proposed approach.
2020 Vol. 33 (4): 363-374 [Abstract] ( 362 ) [HTML 1KB] [ PDF 7693KB] ( 285 )
375 Team Size Optimization for Distributed Patrol of Multi-robot Systems Based on Maximum Idle Time
ZHAO Yuntao, LI Zonggang,DU Yajiang
Aiming at multi-robot patrol problems, a distributed patrol algorithm based on estimated global maximum idleness(EGMI) is proposed to ensure that each patrol vertex can be visited by robots in a certain period of time. In the execution process of algorithm, the global average maximum idle time is estimated using the shared information by each robot, and the next target point to be visited is decided and selected by the robot at the current vertex combining the information collected. Then, performance of the multi-robot team during the patrol task is evaluated according to the global mean maximum idle time. Thus, the optimal number of robots required to complete the patrol task can be obtained. Simulation experiments show that EGMI produces a higher convergence speed and a lower global average maximum idle time. A better result of completing the multi-robot patrol task is achieved.
2020 Vol. 33 (4): 375-382 [Abstract] ( 486 ) [HTML 1KB] [ PDF 714KB] ( 236 )
模式识别与人工智能
 

Supervised by
China Association for Science and Technology
Sponsored by
Chinese Association of Automation
NationalResearchCenter for Intelligent Computing System
Institute of Intelligent Machines, Chinese Academy of Sciences
Published by
Science Press
 
Copyright © 2010 Editorial Office of Pattern Recognition and Artificial Intelligence
Address: No.350 Shushanhu Road, Hefei, Anhui Province, P.R. China Tel: 0551-65591176 Fax:0551-65591176 Email: bjb@iim.ac.cn
Supported by Beijing Magtech 　Email:support@magtech.com.cn