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....
 
 
2016 Vol.29 Issue.8, Published 2016-08-31
Papers and Reports    Researches and Applications   
   
Papers and Reports
673 A Multi-record Webpage Attribute Extraction Method Combining Active Learning
WEI Jingjing, LIAO Xiangwen, CHEN Qiaoling, MA Feixiang, CHEN Guolong
The attribute extraction process can be separated into two phases, alignment and annotation. In the existing alignment methods, different semantic attributes are mistakenly aligned into the same group. Furthermore, to improve the accuracy of semantic annotation, time-consuming manual annotation is often introduced to construct training set. To solve this problem, a multi-record webpage attribute extraction method combining active learning is presented. As for the problem of wrong attribute alignment, shallow semantic is integrated into the alignment approach to relieve the influence of same tags with different semantics. In the semantic annotation phase, textual, visual and global features are extracted for semantic classification and an active learning based SVM classifier is applied to extract structural data. Moreover, a new sample selection strategy is proposed by introducing the global sample information, and more informative samples with lower confidences are selected to be labeled. The experimental results on BBS and microblog datasets confirm the superiority the proposed method.
2016 Vol. 29 (8): 673-681 [Abstract] ( 530 ) [HTML 1KB] [ PDF 603KB] ( 742 )
682 Feature Selection via Evolutionary Computation Based on Relative Classification Information Entropy
ZHAI Junhai, LIU Bo, ZHANG Sufang

Aiming at the problem of feature selection from datasets with discrete values, a feature selection approach via evolutionary computation based on relative classification information entropy is proposed. Genetic algorithm is used to search the optimal feature subset and the relative classification information entropy is employed to measure the significance of the feature subset. Specifically, the relative classification information entropy is used as fitness function, the solutions of the problems are encoded with binary number, and the next generation of individuals is produced by using roulette wheel method. The experimental results show that the proposed approach outperforms other methods in testing accuracy. Furthermore, the proposed approach is theoretically proved to be feasible.
2016 Vol. 29 (8): 682-690 [Abstract] ( 421 ) [HTML 1KB] [ PDF 549KB] ( 680 )
691 Approximation and Reduction Relationships between Multi-granulation Rough Sets and Covering Rough Sets
TAN Anhui, LI Jinjin, WU Weizhi
Multi-granulation rough sets and covering rough sets are two important mechanisms of data processing. From the viewpoint of approximation and reduction, therelationship between multi-granulation rough sets and covering rough sets in complete and incomplete information systems are discussed. Through constructing a space of granules of an information system, it is proved that the optimistic and pessimistic multi-granulation approximations are equivalent to the loose and tense covering approximations, respectively. It means that the optimistic and pessimistic multi-granulation rough sets can be represented by the loose and tense covering rough sets, respectively. Furthermore, two types of consistent sets in multi-granulation rough sets can be transformed into two types of consistent sets in covering rough sets, and there are close relationships of reduction between multi-granulation rough sets and covering rough sets.
2016 Vol. 29 (8): 691-697 [Abstract] ( 433 ) [HTML 1KB] [ PDF 341KB] ( 643 )
698 Method for Social Network User Feature Recognition Based on Clique
HU Kaixian, LIANG Ying, SU Lixin, XU Hongbo, FU Chuan
Social network is a major media for people to get different information and make friends. As the social network keeps developing, it brings convenience to people but meanwhile identifying user identity becomes difficult. To solve this problem, a method for social network user feature recognition based on clique is proposed. According to three degrees of influence rule, the inference model is built, and through the analysis of the clique consisting of user attributes in the social network topology structure, the unknown identity of the current user is inferred. Identity feature recognition methods based on clique are put forward. They are the current user included clique identity recognition method and the multi-degree passing clique identity recognition method. In both methods,the adjacent matrix of social network topology graph of current three-degree friends of user is used to infer the unknown identity of current user via major voting scheme. By the proposed method, the problem of unstable user feature recognition caused by the lack of social relationship is effectively solved. The experimental result shows the good precision of the proposed method.
2016 Vol. 29 (8): 698-708 [Abstract] ( 783 ) [HTML 1KB] [ PDF 596KB] ( 592 )
709 Human Gait Recognition Using Continuous Density Hidden Markov Models
WANG Xiuhui, YAN Ke
As a remote and indirect recognition technology, human gait recognition has extensive applications in various fields, such as video-based surveillance systems. In this paper, the continuous density hidden Markov models (CD-HMM) is employed to perform gait recognition. Firstly, a feature extraction algorithm is proposed based on natural gait cycles,and the observation vector set is constructed using the extracted features. Then, the gait vector set extracted from the training sample set is used to estimate the parameters of CD-HMM. Finally, an adaptive algorithm is introduced based on Cox regression analysis to adaptively adjust parameters of the trained gait model. Experimental results show that the proposed method produces higher accuracies compared with other methods.
2016 Vol. 29 (8): 709-716 [Abstract] ( 650 ) [HTML 1KB] [ PDF 522KB] ( 730 )
717 Feature Joint Probability Distribution and Instance Based Transfer Learning Algorithm
ZHAO Peng, WU Guoqin, LIU Huiting, YAO Sheng
Aiming at the poor generalization ability of only matching marginal probability distribution to reduce the domain difference, a feature joint probability distribution and instance based transfer learning algorithm (FJPD-ITLA) is proposed. The instances are represented with the kernel principal component analysis in subspace. In this subspace, the maximum mean discrepancy is expanded to jointly match the marginal and conditional probability distribution. Thus, the difference between the source domain and target domain is reduced. Meanwhile, the L2,1-norm constraint is utilized to choose relevant instances in the source domain, and the generalization ability of the model obtained by transfer learning is improved further. Experimental results on the digital and object recognition datasets demonstrate the validity and efficiency of the proposed algorithm.
2016 Vol. 29 (8): 717-725 [Abstract] ( 638 ) [HTML 1KB] [ PDF 406KB] ( 1003 )
Researches and Applications
725 Collaborative Filtering Recommendation Algorithm Based on Joint Nonnegative Matrix Factorization
HUANG Bo, YAN Xuanhui, LIN Jianhui
To reveal the hidden relationship between complex network structures and recommend items to users more accurately, an algorithmbased on joint nonnegative matrix factorization (JNMF) is proposed. In the algorithm, user-based collaborative filtering is combined with item-based collaborative filtering. The validity and the convergence of the algorithm are presented in the appendix as well. The experimental results show that the proposed algorithm can combine user-based collaborative filtering algorithm and item-based collaborative filtering algorithm effectively, reduce the mean absolute error to some extent and improve the accuracy of recommendation.
2016 Vol. 29 (8): 725-734 [Abstract] ( 606 ) [HTML 1KB] [ PDF 551KB] ( 936 )
735 Remote Sensing Image Inpainting Based on Non-local Sample Filling and Adaptive Curvature Driven Diffusions Model
WANG Xianghai, SUN Li, WAN Yu, WANG Shuang, TAO Jingzhe
Remote sensing image inpainting technology is significant for the following treatment and application of remote sensing image. Based on a thorough study of curvature driven diffusions (CDD) model and sample filling algorithm, a remote sensing image inpainting algorithm based on non-local sample filling and adaptive curvature driven diffusion model is proposed. The proposed algorithm can avoid the false edge, the staircase effect, the slow diffusion velocity, etc. in some extreme cases during the process of image inpainting. Meanwhile, it maintains the texture feature and edge information well for the inpainted image. The proposed algorithm is verified by simulation experiments.
2016 Vol. 29 (8): 735-743 [Abstract] ( 500 ) [HTML 1KB] [ PDF 6507KB] ( 509 )
744 Robust Kernel-Based Fuzzy Clustering Using Difference of Convex Functions Programming
HE Dan, CHEN Songcan
A nonconvex optimization approach is presented for the robust kernel-based clustering algorithms represented by the radial basis function and the Euler kernel function. The presented approach can handle local optimum problem caused by the non-convexity of the objective function. Difference of convex functions programming (DCP) is applied to escape the local optimum. The clustering accuracy is improved by transforming the objective function into the difference of the two convex functions. The fast and robust algorithm, difference of convex functions algorithm (DCA), is employed to optimize DCP. Consequently, a more robust and optimal solution can be searched by DCP and DCA. Experiments on several UCI datasets show the superiority of the algorithms based on DCP, especially on the large-scale datasets.
2016 Vol. 29 (8): 744-750 [Abstract] ( 521 ) [HTML 1KB] [ PDF 372KB] ( 969 )
751 Sliding Window Prior Knowledge-Based Algorithm for Changepoint Detection in Non-homogeneous Dynamic Bayesian Networks
YU Lu, GAO Yang, SHI Yinghuan
To relax the homogeneity assumption of dynamic bayesian networks (DBNs), the non-homogeneous DBNs is proposed. In this paper, an improved reversible-jump Markov chain Monte Carlo (RJ-MCMC) algorithm is put forward by integrating the prior knowledge about the sliding window, namely APK-RJ-MCMC. The basic assumption of APK-RJ-MCMC is that the bigger the distribution distance between the backward window and the forward window of a time point is, the higher the probability of the time point as a changepoint becomes. Based on the above assumption, the rough probability of each time point as a changepoint is obtained. And it is considered as prior knowledge to guide birth, death and shift moves in RJ-MCMC algorithm during the changepoint sampling. Finally, the accept probability is thus adjusted. Experimental results on both the synthetic data and the real gene expression data show that the proposed APK-RJ-MCMC has a higher posterior probability and better AUC scores than the traditional algorithm does in changepoint detection.
2016 Vol. 29 (8): 751-759 [Abstract] ( 569 ) [HTML 1KB] [ PDF 462KB] ( 551 )
760 Micro-expression Recognition Based on Global Optical Flow Feature
ZHANG Xuange, TIAN Yantao, YAN Fei, WANG Meiqian
The global optical flow feature extraction algorithm based on gradient is studied to improve the effect of micro-expression recognition. To solve the problem of large displacement between fine images, the multi-resolution strategy is introduced to slice the images, and the iterative reweighted least squares method is used to optimize the objective function layer by layer. Thus, the optimal optical flow is obtained, and the accuracy of motion tracking is ensured. To reflect the action differences in key parts of faces, a partition feature statistic method is proposed. The optical flow image is divided into a number of rectangular regions and in these regions the optical flow motion is concluded. Consequently, the effectiveness of the feature is enhanced. The experimental results show that overall recognition accuracy and discrimination of emotion categories are significantly improved.
2016 Vol. 29 (8): 760-768 [Abstract] ( 680 ) [HTML 1KB] [ PDF 653KB] ( 1145 )
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