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....
 
 
2014 Vol.27 Issue.12, Published 2014-12-30
Papers and Reports    Researches and Applications   
   
Papers and Reports
1057 A Context Based ROI Classification Method in Medical Images
GUO Qiao-Jin, LI Ning, XIE Jun-Yuan
Region of interest (ROI) classification is the last and very important step in the process of computer-aided diagnosis with medical images. Traditional methods only employ local visual features of ROI for classification. Thus, the accurate classification can not be achieved under some circumstances. To improve the classification accuracy, the context information is extracted from regions around ROI. A latent Dirichlet allocation classification (LDAC) model based on LDA is proposed, which utilizes LDA to capture contextual information of ROI from surrounding regions. The proposed model is applied to mammograms and experimental results show that the classification accuracy is improved.
2014 Vol. 27 (12): 1057-1064 [Abstract] ( 649 ) [HTML 1KB] [ PDF 740KB] ( 785 )
1065 Attribute Reduction in Variable Precision Rough Set Based on Dependence Space
YU Cheng-Yi, LI Jin-Jin
To get the attribute reduction in variable precision rough set model, an upper and lower approximation binary relation is defined on object sets. By applying the binary relation, the equivalence relation is constructed on attribute sets and thus a dependence space is produced. Then, theorems for judging upper and lower approximation consistent sets are obtained. Meanwhile, a new attribute reduction method is proposed to preserve some invariant characters of upper and lower approximation in each decision class. Finally, a practical example illustrates the validity of the proposed method.
2014 Vol. 27 (12): 1065-1070 [Abstract] ( 432 ) [HTML 1KB] [ PDF 315KB] ( 590 )
1071 Efficient Nearest Neighbor Search Approach for Registration of Low Dimensional Point Sets
ZHU Ji-Hua, YIN Jun, HAN Wen-Xin, DU Shao-Yi
To improve the efficiency of point set registration, an efficient nearest neighbor search approach for 2D/3D point sets is proposed. Firstly, according to the variance of each dimension of the model points, all model points based on the selected dimension information are sorted. By adopting the binary search strategy, each data point is inserted into the sorted model points. Then, the upper bound of search range can be obtained by calculating the distance between the data point and its first left model point. During the search process, the search range can be further reduced by the current nearest neighbor so that the final nearest neighbor can be efficiently searched. Finally, the efficiency of the approach is demonstrated by both the complexity analysis and experimental results comparision.
2014 Vol. 27 (12): 1071-1077 [Abstract] ( 479 ) [HTML 1KB] [ PDF 737KB] ( 605 )
1078 Co-evolutionary Multi-objective Optimization Algorithm with Polymorphous Populations
CHEN Zhen-Xing, YAN Xuan-Hui, WU Kun-An
To improve the diversity maintenance ability of evolutionary multi-objective optimization algorithms and obtain a set of better distributed non-dominated solutions, a co-evolutionary multi-objective optimization algorithm with polymorphous populations is proposed. Firstly, a co-evolutionary frame of polymorphous populations is designed. Next, by introducing the minimum vectorial angle which is capable of measuring the similarity between different Pareto-ranked solutions, a selection strategy for suboptimum non-dominated solutions is proposed to enhance the diversity of populations. Finally, a population removal strategy based on an ordered link-list is put forward. Thus, the uniformity and the spread of the solutions are improved. Compared with some typical algorithms, the proposed algorithm has good convergence and remains a better diversity and uniformity.
2014 Vol. 27 (12): 1078-1088 [Abstract] ( 457 ) [HTML 1KB] [ PDF 944KB] ( 778 )
Researches and Applications
1089 Adaptive Regularization Based Kernel Two Dimensional Discriminant Analysis
JIANG Wei1, ZHANG Jing, YANG Bing-Ru
In traditional semi-supervised dimension reduction techniques, the manifold regularization term is defined in the original feature space. However, its construction is useless in the subsequent classification. In this paper, adaptive regularization based kernel two dimensional discriminant analysis (ARKTDDA) is presented. Firstly, each image matrix is transformed as the product of two orthogonal matrices and a diagonal matrix by using the singular value decomposition method. The column vectors of two orthogonal matrices are transformed into high dimensional space by two kernel functions. Then, the adaptive regularization is defined in the low dimensional feature space, and it is integrated with two dimensional matrix nonlinear method into one single objective function. By altering iterative optimization, the discriminative information is extracted in two kernel subspaces. Finally, experimental results on two face datasets demonstrate that the proposed algorithm obtains considerable improvement in classification accuracy.
2014 Vol. 27 (12): 1089-1097 [Abstract] ( 417 ) [HTML 1KB] [ PDF 593KB] ( 582 )
1098 Automatic Extraction Method for Texture Periodicity Based on Improved Normalized Distance Matching Function
JIANG Sheng, TANG Guo-An, TAO Yang
Based on improved normalized distance matching function (INDMF), an automatic extraction method for regular and near-regular structural texture periodicity is proposed. Firstly, the dissimilarity of gray level co-occurrence matrices is calculated as the texture characteristic, and the INDMF edge is removed. Thus, the values between different peak intervals are more stable. Secondly, an adaptive and anti-noise peak searching approach is adopted to find initial periodic sequence and extract texture periodicity. Next, with the consideration of the characteristics of artificial and natural texture, the final periodicity is calculated by sequence mode. The results of extraction experiments on Brodatz and PSU datasets show the effectiveness and the efficiency of the proposed method. Moreover, the proposed method is more stable and accurate than the method of forward difference of accumulative DMF for impulsive salt and pepper noisy images and projective deformed images.
2014 Vol. 27 (12): 1098-1104 [Abstract] ( 334 ) [HTML 1KB] [ PDF 1439KB] ( 689 )
1105 Multi-tenant Service Customization Algorithm Based on MapReduce and Multi-objective Ant Colony Optimization
WANG Hui-Ying, NI Zhi-Wei, WU Zhang-Jun
Multi-tenant service customization is one of the key technologies to facilitate the agile SaaS multi-tenant architecture, and it can meet the ever-changing personalized demands from customers as well. The hierarchical graph and the customization process of multi-tenant service customization are employed in this paper, and a customization algorithm based on MapReduce and multi-objective ant colony optimization (MSCMA) is proposed. The most suitable business process and the optimized service composition can be found out from various business processes and massive services according to the non-functionality requirement of the tenant, and the optimization tasks can be fulfilled in distributed cloud computing environment in parallel by MSCMA. The results of the simulated experiment demostrate that MSCMA shows favorable convergence and scalability in solving multi-tenant service customization and the proposed algorithm has good ability in processing massive data and solving large scale problems.
2014 Vol. 27 (12): 1105-1116 [Abstract] ( 399 ) [HTML 1KB] [ PDF 714KB] ( 635 )
1117 Multi-robot Pursuit Evasion Based on Quantum Minority Game
WANG Hao, ZHANG Hao, FANG Bao-Fu
When the conflict between interests of pursuers and the overall interests is generated in many-to-one pursuit, more price will be paid by the pursuit system. The classical strategy space can be extended to the range of quantum strategy space after adjusting the payoff distribution mechanism and introducing the quantum minority game. The global optimization can be achieved when robots maximize their own interests in this space. According to the experimental analysis of quantum minority game in the pursuit process, the individual and overall interests are unified,and the robots significantly improve the efficiency by quantum strategy.
2014 Vol. 27 (12): 1117-1123 [Abstract] ( 441 ) [HTML 1KB] [ PDF 508KB] ( 736 )
1124 Regression Analysis for Functional Data Based on Least Squares Support Vector Machine
MENG Yin-Feng, LIANG Ji-Ye
Partial functional linear model is used to explore the relationship between the mixed-type input containing a functional variable and a numerical vector and a numerical output. To improve the accuracy of prediction, based on the representation of the functional coefficient in reproducing kernel Hilbert space, the structured representation of the model is obtained. The estimation problem of the functional coefficient is converted into the estimation problem of a parameter vector, and the least squares support vector machine method is used for parameter estimation. Experimental results show that the performance of vector coefficient estimator is similar to other parameter estimation methods while the functional coefficient estimator is stabler and more accurate than the others, and the good performance of the proposed method further ensures the accuracy of machine learning.
2014 Vol. 27 (12): 1124-1130 [Abstract] ( 544 ) [HTML 1KB] [ PDF 490KB] ( 1153 )
1131 Feature Selection Algorithm for Incomplete Data Based on Information Entropy
CHEN Sheng-Bing, WANG Xiao-Feng
Grounded on the analysis of the existing incomplete information entropy, the concept of incomplete information entropy based on similarity relations (SIIE) is proposed, and some properties of SIIE are discussed. A feature selection algorithm for incomplete data is presented. In this algorithm, SIIE of incomplete data is calculated directly, and SIIE is taken as the criteria for feature selection. Then, the sequential forward floating search method is employed to addresses the problem of correlation among features. Experiments on UCI database are carried out, and the results indicate the accuracy and efficiency of the proposed algorithm.
2014 Vol. 27 (12): 1131-1137 [Abstract] ( 588 ) [HTML 1KB] [ PDF 501KB] ( 917 )
1138 Application of Imbalanced Data Learning Algorithms to Similarity Learning
XIA Pei-Pei, ZHANG Li

In the real-world problems, there is an imbalance in the paired-samples. The number of the paired-samples in similarity set is much smaller than the number of the paired-samples in dissimilarity set. To solve this problem, two approaches, dissimilar K nearest neighbor and similar K nearest neighbor (DKNN-SKNN) and dissimilar K nearest neighbor and similar K farthest neighbor (DKNN-SKFN), are proposed to construct paired-samples. Thus, the number of paired-samples in similarity learning is effectively decreased, the training process of SVM is accelerated, and the imbalanced data problem is solved to some degree. In the experiments, the proposed approaches are compared with some standard resampling methods. The results show that the proposed approaches have better performance.
2014 Vol. 27 (12): 1138-1146 [Abstract] ( 391 ) [HTML 1KB] [ PDF 582KB] ( 777 )
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