Home      About Journal      Editorial Board      Instructions      Ethics Statement      Contact Us                   中文

Judgement and Disposal of Academic Misconduct Article Copyright Transfer Agreement Proof of Confidentiality Requirements for Electronic Version More.... Chinese Association of Automation National ResearchCenter for Intelligent Computing System Institute of Intelligent Machines,Chinese Academy of Sciences More....

2011 Vol.24 Issue.1, Published 2011-02-01 Orignal Article        Orignal Article 1 Construction of Complete Description Logics Ontologies Using Attribute Exploration TANG Su-Qin, CAI Zi-Xing, WANG Ju, JIANG Yun-Cheng The importance and current research progress of the description logics ontologies are analyzed, especially the application of attribute exploration in the construction of description logics ontologies according to the completeness of the description logics ontologies. The insufficiency in the knowledge that domain experts are supposed to have in using attribute exploration to construct the description logics ontologies is discussed. The construction of complete description logics ontologies under which circumstances domain experts do not have all the knowledge required in the domain is discussed as well. Moreover, a definition for the completeness of description logics ontologies under the description contexts is provided, and the incomplete contexts under the description contexts are set. An algorithm of constructing the description logics ontologies is constructed under the condition that domain experts are unable to define the attribute implications between those attribute sets. The proposed algorithm is used to acquire an embodying knowledge and then construct the knowledge base. And it can be proved that the description logics ontologies constructed by the proposed method is a complete one. 2011 Vol. 24 (1): 1-13 [Abstract] ( 361 ) [HTML 1KB] [ PDF 950KB] ( 988 ) 14 Symbolic ADD Algorithms for Weighted Constraint Satisfaction Problem XU Zhou-Bo, GU Tian-Long , CHANG Liang The weighted constraint satisfaction problem (WCSP) is a kind of soft constraint satisfaction problems with many practical applications. An algebraic decision diagram (ADD) based scheme is proposed to solve WCSP efficiently. Firstly, the soft constraint network is represented as ADDs so that the WCSP can be formulated symbolically and manipulated implicitly. Secondly, the symbolic ADD algorithm is presented, where the branch and bound algorithm is integrated with bucket elimination algorithm symbolically. And the static variable orderings and the node consistency during search are used. To improve the lower bound of the symbolic ADD algorithm, the counting technique of directional arc consistency is adopted, and thus the symbolic ADD algorithm is improved. Finally, experiments on random problems are implemented, and the results show that the proposed algorithms outperform the depth-first branch and bound algorithms enhanced with a look-ahead process and a preprocessing step of either existential directional arc consistency or the node consistency. 2011 Vol. 24 (1): 14-21 [Abstract] ( 347 ) [HTML 1KB] [ PDF 501KB] ( 739 ) 22 Fast Twin Support Vector Regression Algorithm in Primal Space PENG Xin-Jun, WANG Yi-Fei Twin support vector regression (TSVR) efficiently determines its objective regression function by optimizing a pair of smaller sized SVM-type problems. The objective functions of TSVR in the primal space are directly optimized by introducing the well-known Newton algorithm. This method effectively overcomes the shortcoming of TSVR that its regressor is approximated by the dual quadratic programming problems. Numerical studies show that the proposed method provides good performance and obtains less learning time compared with TSVR. 2011 Vol. 24 (1): 22-29 [Abstract] ( 389 ) [HTML 1KB] [ PDF 399KB] ( 693 ) 30 Crowd Density Classification Based on Confidence Analysis MA Wen-Hua , HUANG Lei, LIU Chang-Ping Crowd density estimation is crucial for crowd monitoring and is mainly used for calculating quantified levels for crowd density of target monitor areas in videos or images. A crowd density classifier is proposed based on confidence analysis. Several binary classifiers are firstly combined together by error correcting output codes, which is designed under the guidance of binary tree theory. Confidence samples are selected and used for training support vector machines, which are adopted as binary classifiers. The decoding algorithm is based on transmission channel model and the samples are assigned to classes with maximum posterior probabilities. Experimental results demonstrate that the proposed approach is superior to the traditional classification models under the premise of same dataset and features, which provides a method for multi-category classification such as crowd density estimation. 2011 Vol. 24 (1): 30-39 [Abstract] ( 469 ) [HTML 1KB] [ PDF 724KB] ( 934 ) 40 CAPTCHA Recognition Method Based on RNN of LSTM ZHANG Liang, HUANG Shu-Guang, SHI Zhao-Xiang, HU Rong-Gui Completely automated public turing test to tell computers and humans apart (CAPTCHA) is a kind of network security mechanism based on hard artificial problems. Study of recognition of CAPTCHA impels it to become more secure, and some hard atifical problems to be solved. Firstly, CAPTCHA recognition methods of state of the art are analyzed. Then, a recognition method is brought up based on recurrent neural network (RNN) which is composed by long short-term memory (LSTM) blocks. Thirdly, feature extraction for CAPTCHA recognition is studied. Finally, a decoding algorithm is proposed to improve the recognition rate. Experimental results show that the proposed recognition method is efficient. Gray value of images is proved to be a kind of good feature for RNN. Furthermore, the proposed decoding algorithm gets high recognition rates with low time complexity. 2011 Vol. 24 (1): 40-47 [Abstract] ( 1210 ) [HTML 1KB] [ PDF 576KB] ( 20469 ) 48 Semi-Supervised Eigenvector Selection for Spectral Clustering ZHAO Feng, JIAO Li-Cheng, LIU Han-Qiang, GONG Mao-Guo For a K clustering problem, Ng-Jordan-Weiss (NJW) spectral clustering method adopts the eigenvectors corresponding to the K largest eigenvalues of the normalized affinity matrix derived from a dataset as a novel representation of the original data. However, these K eigenvectors can not always reflect the structure of the original data for some pattern recognition problems. In this paper, a semi-supervised eigenvector selection method for spectral clustering is proposed. This method utilizes some amount of supervised information to search the eigenvector combination which can reflect the structure of the original data, and then obtains more satisfying performance than the classical spectral clustering algorithms. Experimental results on UCI benchmark datasets and MNIST handwritten digits datasets show that the proposed method is effective and robust. 2011 Vol. 24 (1): 48-56 [Abstract] ( 526 ) [HTML 1KB] [ PDF 675KB] ( 1044 ) 57 Object Tacking with Particle Filter Based on Multi-Agent Co-Evolution LI Yong-Ping, WANG Yan-Jiang, QI Yu-Juan When tracking the moving targets in video image sequences, the existing particle filter is not satisfactory due to the particle degradation and particle diversity loss. An object tracking algorithm with particle filter is proposed. The multi-agent co-evolutionary mechanism is introduced into the particle resampling process and makes the particle become an agent with abilities of local perception, competitive selection and self-learning by redefining the particle agent and its local living environment. The resampling process is accomplished by the co-evolutionary behaviors among particles such as competition, crossover, mutation and self-learning. It ensures the particle validity and increases the particle diversity. Experimental results show that the proposed algorithm tracks the moving object accurately and robustly in complex video scenes. 2011 Vol. 24 (1): 57-63 [Abstract] ( 431 ) [HTML 1KB] [ PDF 732KB] ( 763 ) 64 A Survey of Evaluation and Design for AUC Based Classifier WANG Yun-Yun, CHEN Song-Can Though as a common performance evaluating index for classification algorithms, accuracy (or total misclassification error) has several deficiencies, such as the sensitivity to class prior distribution and misclassification costs, and the ignorance of the posterior probability and ranking information obtained by classification algorithms. While the area under the receiver operation characteristic (ROC) curve measures the classification performance across the entire range of class prior distribution and misclassification costs, as well as the probability and ranking performance. Thus, it attracts much attention in classification learning and evokes a lot of researches. In this paper, a relative comprehensive survey for these researches is presented, including the advantages of AUC as a performance evaluating index, the design of algorithms based on AUC, the relationship between the accuracy-maximizing and AUC-maximizing algorithms and the deficiencies of AUC along with its variants. 2011 Vol. 24 (1): 64-71 [Abstract] ( 372 ) [HTML 1KB] [ PDF 621KB] ( 955 ) 72 Adaptive Sequence Learning and Applications for Multi-Scale Kernel Method WANG Hong-Qiao, CAI Yan-Ning, SUN Fu-Chun, ZHAO Zong-Tao Multi-scale kernel method is a hotspot of current kernel machine learning field. However, in the multiple kernel processing progress of multi-scale kernel learning methods, there are some disadvantages, such as average combination of kernels, time consumption increasing under iterative training and empirical selection of composite coefficients. Based on the kernel target alignment heuristics, an adaptive sequence learning algorithm for multi-scale kernel method is presented and the weighting coefficients of multiple kernels can be obtained automatically and rapidly. The experimental results testify that the proposed algorithm has better performance and stability in regression precision and classification accuracy than the SVM methods using different single kernels. Moreover, the proposed algorithm has good universal applicability. 2011 Vol. 24 (1): 72-81 [Abstract] ( 576 ) [HTML 1KB] [ PDF 661KB] ( 945 ) 82 Recognition of Online Multistroke Freehand Conic Sections Based on Time-Space Relationship WANG Shu-Xia, WANG Guan-Feng, GAO Man-Tun, YU Sui-Huai A method for interpreting multistroke freehand conic sections is presented based on the time-space relationship. The pretreatment approach based on the time slot is able to interpret the dashed or continuous strokes, and the recognition method is discussed based on space relationship of strokes. The multistroke interpretation of freehand conic sections is divided into the interpretations of the closed and non-closed conic sections. Using the least median squares method fits conic section for the former. The definitions of the rotation angle and the endpoint of generalized stroke of the non-closed conic section are presented, which consists of three stages: the multistroke determination, the sketch generation and the rotation angle and endpoints calculation. A human-computer interface prototype system FSR is developed using the proposed theory, which makes system interface easy and friendly to user. The FSR system is tested with a number of multistroke sketches and the results show that the proposed algorithm achieves a satisfactory interpretative efficiency. 2011 Vol. 24 (1): 82-89 [Abstract] ( 377 ) [HTML 1KB] [ PDF 558KB] ( 693 ) 90 Cloud Hypermutation Particle Swarm Optimization Algorithm Based on Cloud Model ZHANG Ying-Jie, SHAO Sui-Feng , Niyongabo Julius Integrated with the basic principle of particle swarm optimization, a rapid evolutionary algorithm is proposed based on the characteristics of the cloud model on the process of transforming a qualitative concept to a set of quantitative numerical values, namely cloud hypermutation particle swarm optimization algorithm. Its core idea is to achieve the evolution of the learning process and the mutation operation by the normal cloud particle operator. With the cloud model, inheritance and mutation of the particle can be modeled naturally and uniformly, which makes it easy and nature to control the scale of the searching space. The simulation results show that the proposed algorithm has fine capability of finding global optimum, especially for multimodal function. 2011 Vol. 24 (1): 90-96 [Abstract] ( 384 ) [HTML 1KB] [ PDF 387KB] ( 625 ) 97 Active Learning Algorithm Based on Neighborhood Entropy WANG Zhen-Yu, WANG Xi-Zhao Neighborhood entropy is adopted as the sample selection criteria in active learning. The example with the highest entropy value is considered as the most uncertain one based on current nearest neighbor rule. And labeling the most uncertain example can achieve higher accuracy with fewer samples. An active learning algorithm based on neighborhood entropy is proposed. The scheme estimates entropy value of neighbor unlabeled sample and label the sample with the highest value. Experimental results show the example selection based on neighborhood entropy achieves higher accuracy compared with maximal distance sampling and random sampling. 2011 Vol. 24 (1): 97-102 [Abstract] ( 387 ) [HTML 1KB] [ PDF 367KB] ( 606 ) 103 Text Feature Selection Method for Hierarchical Classification ZHU Cui-Ling, MA Jun, ZHANG Dong-Mei An approach of feature selection for hierarchical classification is proposed. Firstly, the concept of category hierarchical correlation degree is introduced and it is calculated according to the category tree and the probability distribution of training data on different levels. Then, the importance degrees of categories are computed according to hierarchical correlation degree. Finally, the discriminative abilities of features are calculated based on the previous computation and the features with the greater discriminative ability are chosen as the feature set for classification. Experimental results show that the proposed approach outperforms the traditional feature selection methods on both quality of the features selected and standard classification metrics in terms of accuracy, F1 and micro-precision. 2011 Vol. 24 (1): 103-110 [Abstract] ( 444 ) [HTML 1KB] [ PDF 515KB] ( 702 ) 111 Semantic Web Session Clustering and Visualization Method Based on Ontology YANG Qian-Wen, KOU Ji-Song, CHEN Fu-Zan, LI Min-Qiang A representation method for generalized web sessions based on ontology is proposed, called semantic web session, and the corresponding semantic web session clustering and visualization method is presented. For the sessions clustering, the similarity measure on semantic common paths (SMSCP) for semantic web sessions is defined on the semantic common paths of users navigation. The validity of the similarity measure is verified through the clustering accuracy using the improved k medoids algorithm. Stratograms are employed to visualize the clustering results. Experimental results show that the proposed clustering and visualization methods are effective and understandable. 2011 Vol. 24 (1): 111-116 [Abstract] ( 371 ) [HTML 1KB] [ PDF 385KB] ( 598 ) 117 Multiscale Registration Based on Edge-Preserved Scale Space for Medical Images LI Deng-Wang, WANG Hong-Jun, YIN Yong The limitation of the conventional multiresolution registration framework is analyzed from the perspective of scale space filtering. Edge-preserved scale space is proposed for multi-scale registration to improve the accuracy and speed and avoid local extreme. The proposed framework has a good edge preserved property which provides more spatial information for mutual information based registration. To achieve automatic registration, a method is proposed to obtain the smoothing parameter λ for non-linear diffusion model. The experimental results show that the proposed framework is superior to other traditional frameworks and suitable for 3-D medical image registration. The registration results have higher accuracy with less numbers of iteration. Furthermore, when traditional frameworks fail to register the images, the proposed framework still has accurate registration results, thus the proposed framework has better robustness. 2011 Vol. 24 (1): 117-122 [Abstract] ( 371 ) [HTML 1KB] [ PDF 443KB] ( 836 ) 123 Ambiguity Reduction Based on Qualitative Mutual Information in Qualitative Probabilistic Networks L Ya-Li, LIAO Shi-Zhong To reduce the inference ambiguity in the sign-propagation algorithm, a method is proposed based on qualitative mutual information in qualitative probabilistic networks (QPN). Firstly, the definition of qualitative mutual information is given. Then, an enhanced formalism of qualitative probabilistic networks (EQPN) is presented based on this definition, which can distinguish between strong and weak influences. Thirdly, symmetry, transitivity and parallel composition of qualitative influences in EQPN are analyzed. Finally, the correctness and efficiency of the sign-propagation algorithm in EQPN are verified by experiments on the Antibiotics database. Theoretic analysis and experimental results show that EQPN is qualitative, efficient, and it reduces inference ambiguity correctly. 2011 Vol. 24 (1): 123-129 [Abstract] ( 271 ) [HTML 1KB] [ PDF 425KB] ( 702 ) 130 Data Extraction from Limited Deep Web Based on Latticial Space ZHANG Zhuo, LI Shi-Jun, ZHANG Nai-Zhou, TIAN Jian-Wei In the situation of crawling Deep Web database that limits the number of results, the problem of appropriately predicting the results size of queries can be modeled as a set covering problem with condition of limited set size. This problem is modeled as a concept covering problem. Firstly, the relation among all couples composed by a query and its result is proved as tolerance. Secondly, set of them is proved as a complete lattice which is homomorphism to the concept lattice from the same source. Therefore, the order relation between concepts can be utilized to describe correlation between queries. The intent of a concept can be considered as a query, thus the result size is forecasted by cardinality of the concept extent. A lattice-based algorithm is proposed for data extraction from limited Deep Web database, called Ladeldew. Semi-lattice pruned based on the cardinality of extent is exploited by Ladeldew as search space. The new search space is iteratively generated from new data until nothing can be extracted. Both controlled and real experiments are implemented to evaluate Ladeldew, and the results verify its theoretical correction and realistic application. 2011 Vol. 24 (1): 130-137 [Abstract] ( 327 ) [HTML 1KB] [ PDF 542KB] ( 648 ) 138 OCPA Bionic Autonomous Learning System and Its Application to Robot Poster Balance Control CAI Jian-Xian, RUAN Xiao-Gang An operant conditioning probabilistic automation (OCPA) bionic autonomous learning system is constructed according to nonlinear, strong-coupling and complex two-wheeled self-balancing robot dynamic system. The OCPA bionic autonomous learning system is a probabilistic automaton based on Skinner operant conditioning whose main character lies in simulating the operant conditioning mechanism of biology. And it has bionic self-organization function which contains the self-learning and adaptive functions, and thus the OCPA automaton can be used to describe, simulate and design various self-organization systems. The convergence of operant conditioning learning algorithm of OCPA learning system is proved theoretically. The results of both simulation and experiment applied to two-wheeled robot poster balance control indicate that the OCPA learning system does not require the robot model, and the motion balanced skills of robot are formed, developed and perfected gradually by simulating the operant conditioning mechanism of biology. 2011 Vol. 24 (1): 138-146 [Abstract] ( 356 ) [HTML 1KB] [ PDF 621KB] ( 1580 ) 147 Open Set Face Recognition Approach Based on Similarity Distribution ZHANG Kai, SU Jian-Bo A classification algorithm based on multi-dimensional similarity distribution is presented to enhance the accuracy in open set face recognition. This algorithm firstly get the similarity vector distribution of known and unknown samples by testing on many labeled pictures. Then those similarity vectors are learned by linear discriminant analysis (LDA) to extract distribution features. Finally, the proposed algorithm rejects the unknown identity by feature-matching. Hence, the feature has strong classification ability in view of the discrimination information abstracted from the similarity distribution. Experimental results on several face databases demonstrate that the proposed method significantly outperforms the traditional method for open set face recognition. 2011 Vol. 24 (1): 147-152 [Abstract] ( 488 ) [HTML 1KB] [ PDF 395KB] ( 789 )

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