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....

2013 Vol.26 Issue.9, Published 2013-09-30 Orignal Article        Orignal Article 801 A Collaborative Solving Algorithm for Dynamic Distributed Constraint Optimization Problem GE Fang-Zhen,WEI Zhen,LU Yang,QIU Shu-Wei,LI Li-Xiang A large number of problems in the multiagent collaboration process can be modeled under the framework of distributed constraint optimization problem (DCOP). However,DCOP framework is limited to the issue of planning,and the agents in DCOP generally require a complete and accurate reward function. To resolve this issue,a dynamic distributed constraint optimization problem (DDCOP) is defined,and DDCOP′s crucial operations,exploration and exploitation,are analyzed. Furthermore,a chaotic ant based collaborative solving algorithm for dynamic distributed constraint optimization problem (CA-DDCOP) is proposed. The CA-DDCOP algorithm is established based on chaotic behavior of a single ant and self-organizing behavior of ant colony,thereby the exploration and exploitation are realized. The proposed algorithm achieves the collaboration of exploration and exploitation according to the Boltzmann distribution. Then a channel allocation in multi-radio multi-channel Ad Hoc networks is solved by the CA-DDCOP algorithm. The simulation results show that the CA-DDCOP algorithm performs effectively. 2013 Vol. 26 (9): 801-811 [Abstract] ( 463 ) [HTML 1KB] [ PDF 720KB] ( 1412 ) 812 A 2D+3D Multimodal Ear Recognition Method YUAN Li,MU Zhi-Chun A 2D+3D multimodal ear recognition method is proposed. Firstly,ear detection method based on Adaboost algorithm is used to detect ear part on the 2D images,then the corresponding ear part is located and extracted in the 3D range image. For 2D ear recognition,Kernel Fisher Discriminant Analysis is applied for feature extraction and Nearest Neighbor classifier is applied for ear recognition. For 3D ear recognition,3D Local Binary Pattern descriptor is applied for feature extraction on range image,geometric constraint and location constraint are used to perform the matching process between a test ear and a registered protocol ear,and ear recognition performance is evaluated by the number of the matching points. Finally,Bayes decision rule is used for the decision level fusion of 2D and 3D ear recognition classifiers. The experimental results on the UND ear dataset show the effectiveness of the proposed method. In lighting variation scenario,the proposed 2D+3D fusion method outperforms unimodal ear recognition method with 2D images or range images. 2013 Vol. 26 (9): 812-818 [Abstract] ( 471 ) [HTML 1KB] [ PDF 827KB] ( 885 ) 819 Multi-Gene Expression Programming with Depth-First Decoding Principle DENG Wei,HE Pei,QIAN Jun-Yan Gene expression programming (GEP) is an automatic programming approach which is widely used in many areas. As far as the decoding method is concerned,it uses the breadth-first principle to transform individuals into expressions. It means that the meaning of a gene segment will change with the context. Consequently,any individual can not be concurrently evaluated in most existing GEPs. In this paper,the theoretical analysis and experiments show that the depth-first principle as well as multi-solution techniques,i.e. techniques for encoding of multiple solutions into a single chromosome,can not only solve the mentioned GEP problem,but also significantly improve its performance. 2013 Vol. 26 (9): 819-828 [Abstract] ( 437 ) [HTML 1KB] [ PDF 857KB] ( 705 ) 829 A Bat Algorithm Based on Lévy Flights Trajectory XIE Jian,ZHOU Yong-Quan,CHEN Huan Aiming at the phenomenon that bat algorithm has slow convergence and low precision,an improved bat algorithm based on Lévy flights trajectory is proposed. The proposed algorithm is characterized by quick convergence and high precision,and it can effectively jump out of local optimum. By means of 12 typical test functions simulation,the results show that the algorithm is effective and feasible. Moreover,the algorithm also has excellent approximation performance in solving an optimization problem with high-dimensional space. 2013 Vol. 26 (9): 829-837 [Abstract] ( 559 ) [HTML 1KB] [ PDF 843KB] ( 793 ) 838 Heuristically Accelerated State Backtracking Q-Learning Based on Cost Analysis FANG Min,LI Hao Since action strategy learning is time-consuming for the reinforcement learning algorithm,a heuristic reinforcement learning algorithm is presented based on state backtracking. By analyzing the repetitive states and comparing the action policies of the reinforcement learning,a cost function is defined to indicate the importance of repetitive actions. A probability-based heuristic function is presented by combining an action reward with an action cost. The proposed algorithm reinforces the importance of an action to speed up learning by the heuristic function and measures the feasibility of an action to reduce unnecessary exploration by the cost function at the same time,thus the learning efficiency is steadily improve. This cost-based action strategy is proved to be reasonable. Two simulation scenarios are built and the experimental results of robot games prove that the proposed algorithm can learn by the tradeoff between rewards and costs,and effectively improve the convergence of Q-learning. 2013 Vol. 26 (9): 838-844 [Abstract] ( 358 ) [HTML 1KB] [ PDF 495KB] ( 654 ) 845 Unsupervised Uyghur Segmentation and Unsupervised Feature Selection TOHTI Turdi,PATTA Akbarr,HAMDULLA Askar Commonly used Uyghur segmentation method produces a large number of semantic abstraction and even polysemous word features,so learning algorithms are difficult to find the hidden structure in the high-dimensional data. A segmentation approach dme-TS and a feature selection approach UMRMR-UFS based on unsupervised strategy are proposed. In dme-TS,the word based Bi-gram and contextual information are derived from large scale raw text corpus automatically,and the liner combinations of difference of t-test,mutual information and entropy of double word adjacency are taken as a measurement (dme) to estimate the agglutinative strength between two adjacent Uyghur words. In UMRMR-UFS,an improved unsupervised feature selection criterion (UMRMR) is proposed and the importance of each feature is estimated according to its minimum redundancy and maximum relevancy. The experimental result shows that dme-TS effectively reduces the dimensions of original feature set and improves the quality of the feature itself,and the learning algorithm represents its highest performance on the feature subset selected by UMRMR-UFS. 2013 Vol. 26 (9): 845-852 [Abstract] ( 398 ) [HTML 1KB] [ PDF 544KB] ( 834 ) 853 Conflict Evidence Fusion Method Based on Conflict Representation ZHANG Yan-Jun,LONG Cheng,LI Da In order to effectively solve the conflict evidence combination problems,an evidence combination method is proposed. Based on a new conflict representation and the association of quantitative evidence,the weight of each evidence is calculated. Then compared with average weight,the conflict evidence or low credibility of evidence is identified and the discount correction is made,and the use of the original evidence source information is maximized. Finally,the D-S combination rule is employed. The experimental results show that the proposed method has good performance,fast convergence and high reliability compared with other typical methods. 2013 Vol. 26 (9): 853-858 [Abstract] ( 369 ) [HTML 1KB] [ PDF 361KB] ( 712 ) 859 A μ-AEA Constraint Optimization Algorithm Based on AEA WANG Zhen,LI Shao-Jun A constrained handling method based on the Alopex-based evolutionary algorithm(AEA) is proposed. The relatively feasible region is gradually converged to the feasible region by the introducing adaptive relaxation parameter μ in the iteration,which takes into account that different functions have different sizes of feasible regions. Also the relaxation of constraints allows more infeasible individuals which contain some useful information to keep staying in the next generation. And therefore it enhances search ability of the algorithm. At the same time,an adaptive penalty function method is introduced,and it adaptively adjusts the penalty coefficient based on the different constraint satisfactions. Thus,it ensures that the punishment is not too large or too small. 11 standard test function experiments show that the proposed method has satisfactory results and great potential in handling works with constraint optimization problems. 2013 Vol. 26 (9): 859-864 [Abstract] ( 333 ) [HTML 1KB] [ PDF 344KB] ( 453 ) 865 Parameter-Free Locality Preserving Projections and Face Recognition HUANG Pu,TANG Zhen-Min Locality Preserving Projections (LPP) aims to preserve local structure of the data by constructing a nearest-neighbor graph. However,it is confronted with the difficulty of parameter selection in the process of graph construction. To solve this problem,an algorithm called parameter-free locality preserving projections (PLPP) is proposed. Firstly,a parameter-free graph construction strategy is designed,which can actively determine neighbors of each data point and assign corresponding edge weights. Then,with the proposed graph construction strategy,PLPP seeks an optimal transformation matrix to preserve local structure of the data in the low dimensional space. Since PLPP needs no parameters in graph construction and takes cosine distance as the similarity weight,it is more efficient and robust to outliers than LPP. Moreover,supervised PLPP (SPLPP) is proposed to improve the discriminant ability of PLPP by considering class information of samples. The experimental results on the ORL,FERET and AR face databases validate the effectiveness of PLPP and SPLPP. 2013 Vol. 26 (9): 865-871 [Abstract] ( 425 ) [HTML 1KB] [ PDF 462KB] ( 734 ) 872 Feature Extraction Method Based on Global Binary Patternand Its Application XU Ke,SONG Chang Texture analysis based on Global Binary Pattern (GBP) is proposed to solve the problem that Local Binary Pattern (LBP) is sensitive to noises. In GBP,center pixels used in LBP are replaced by the mean values of large neighborhood templates,and effects of noises are weakened. However,the resistance to uneven illumination by GBP is worse than that by LBP. For surface defect recognition of steel plates,both noises and uneven illumination are serious in the images of steels. The combination of GBP and LBP with bivariate histogram is presented and applied to surface defect recognition of steel plates and slabs. The experimental results show that the combination of GBP and LBP is invariant to uneven illumination and insensitive to noises,and classification rate of cracks is up to 96%. 2013 Vol. 26 (9): 872-877 [Abstract] ( 418 ) [HTML 1KB] [ PDF 663KB] ( 857 ) 878 Clonal Selection Algorithm with GEP Code for Function Modeling MO Hai-Fang,LI Kang-Shun The clonal selection algorithm evolves through selecting best individuals,cloning the selected ones and hypermutation. The general method to find the best individuals is to sort the individuals according to their fitness. However,the GEP codes of those chromosomes with same fitness may be different. If duplicate individuals are allowed to appear in the sorted population,the duplicate superior individuals will be cloned excessively. In this case,the diversity of the population is decreased. If individuals are sorted just according to their fitness,the duplicate ones will be removed. And some best individuals with different codes may be abandoned. In order to maintain the diversity of population and increase the convergence rate,an improved clonal selection algorithm is proposed. Firstly,the individuals are sorted according to their fitness. Then,if there are multiple best individuals with same fitness,their codes are compared. The best individuals with different codes will be selected to clone. The experimental result shows that the proposed method maintains the diversity of population and increases the convergence rate. 2013 Vol. 26 (9): 878-884 [Abstract] ( 335 ) [HTML 1KB] [ PDF 433KB] ( 595 ) 885 Mixture Particle PHD Filter Based Multi-Target Visual Tracking LIN Qing,XU Xiao-Gang,ZHAN Yong-Zhao,LIAO Ding-An,YANG Ya-Ping Aiming at the problem that particle filter is poor at consistently maintaining the multi-modality of the target distributions for multi-targets in a variable number of visual tracking,a multi-target visual tracking approach based on mixture particle probability hypothesis density (PHD) filter is proposed. The particles are clustered by the K-means algorithm,the classified particles are labeled and the particle filters are separately used for each classified particles. It improves the accuracy of target states estimation and effectively maintains the multi-modal distribution of the various objectives. The experimental results show that the proposed approach is an effective solution to the appearance,merger,separation and other multi-target tracking problems for the new target. 2013 Vol. 26 (9): 885-890 [Abstract] ( 420 ) [HTML 1KB] [ PDF 853KB] ( 582 ) 891 Aggregation of Satisfactory Rankings for CP-Nets Based on Voting System SUN Xue-Jiao,LIU Jing-Lei The satisfactory ranking for conditional preference networks (CP-Nets) is a ranking of all outcomes acquired from preferences of decision maker,and its preference relation is incomplete and transitive. The feasibility on aggregation of satisfactory rankings for CP-Nets with traditional voting system is analyzed. The detailed realizations and performance analysis on aggregation of satisfactory rankings with the majority rule and positional voting are individually introduced. So it realizes the promotion and application of voting system from completeness to non-completeness. 2013 Vol. 26 (9): 891-898 [Abstract] ( 345 ) [HTML 1KB] [ PDF 372KB] ( 547 )

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