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.4, Published 2014-04-30
Papers and Reports    Researches and Applications    Surveys and Reviews   
   
Papers and Reports
289 Performance Evaluation System in Open Set Face Recognition
LIANG Yi-Cong, DING Xiao-Qing, FANG Chi
A performance evaluation system is built for the open set face recognition task. The system consists of two parts: the quality evaluation for training set and the performance evaluation for classification result of test samples. For the former, Bhattacharyya distance is used to approximate the Bayesian error rate, and the particularity of open set problems that sample pairs do not obey the independent identical distribution assumptions is taken into account. The quality evaluation functions of the training set are obtained in both Gaussian or non-Gaussian distribution assumptions, and in Gaussian case this function has a closed form. For the latter, the distribution densities of the nearby positive and negative sample pairs are considered to measure the reliability of the similarity score given by a classifier. Therefore, the previous studies which are lacking of such measurements are complemented. The results in this paper are validated by experiments on multiple face databases.
2014 Vol. 27 (4): 289-293 [Abstract] ( 696 ) [HTML 1KB] [ PDF 810KB] ( 1129 )
294 Face Recognition via Compressive Sensing Based on Fisher Discrimination Dictionary Learning
ZENG Ling-Zi, YIN Dong, ZHANG Rong, ZHEN Hai-Yang
Sparse representation based classification (SRC) algorithm loses much discriminative information hidden in the training samples when constructing dictionary and the L1-minimization approach to solving the coding coefficient is computationally expensive. Aiming at these problems,a face recognition algorithm via compressive sensing based on Fisher discrimination dictionary learning and least square method is proposed. The training samples are trained by Fisher discrimination criterion and thus the structured dictionary is acquired. Then, the coding coefficients are obtained by solving L2-minimization problem through regularized least square method. Finally, the face is identified through the coding coefficient and reconstruction error. The experimental results clearly show that the proposed method has a better accuracy rate and improves the recognition speed compared with the existing sparse representation classification methods.
2014 Vol. 27 (4): 294-299 [Abstract] ( 546 ) [HTML 1KB] [ PDF 498KB] ( 1140 )
300 Upper Approximation Reduction in Intuitionistic Fuzzy Object Information Systems with Dominance Relations
WU Lei, YANG Shan-Lin, GUO Qing
The traditional rough set theory can not be directly used to reduce the attributes of intuitionistic fuzzy object information systems(IFOIS) in which the decision attribute values are intuitionistic fuzzy numbers. In this paper, the dominance relation is introduced to intuitionistic fuzzy object information systems. Based on dominance relation, a new definition of the upper approximation decision consistent set of the condition attribute sets is presented and the judgment theorem of the upper approximation reduction is also given. Thus, the upper approximation reduction model of the condition attribute sets is established. Moreover, an algorithm to compute the upper approximation reduction is put forward. In some object information systems in which decision attribute values are intuitionistic fuzzy numbers, more concise decision rules can be obtained via the upper approximation reduction of the condition attribute sets. Finally, an example is given to illustrate the effectiveness of the proposed algorithm.
2014 Vol. 27 (4): 300-304 [Abstract] ( 599 ) [HTML 1KB] [ PDF 307KB] ( 626 )
305 Self-adapted Harmony Search Algorithm with Opposed Competition and Its Optimization
OUYANG Hai-Bin, GAO Li-Qun1, KONG Xian-Yong, ZOU De-Xuan
A self-adapted harmony search algorithm with opposed competition (SHSOC) is proposed. The blindness of bandwidth setting of harmony search algorithm is analyzed.The adaptive bandwidth adjustment is employed. Meanwhile, the superiority of the opposed learning strategy is integrated into the proposed algorithm, and the competition selection mechanism of end elimination is established to further improve global search ability and keep the algorithm from falling into local optima. The proposed algorithm is tested on several classic functions to evaluate the performance. The numerical results show the superiority of SHSOC in accuracy and robustness compared with harmony search algorithm and some state-of-the-art harmony search variants. Moreover, SHSOC can solve the optimization problems of the heat exchanger and the speed reducer design, and the results show that SHSOC is better than any other algorithm.
2014 Vol. 27 (4): 305-312 [Abstract] ( 352 ) [HTML 1KB] [ PDF 579KB] ( 1062 )
313 Evidential Reasoning Approach for Solving Complex Evaluation Models
FU Yang-Geng, YANG Long-Hao, WU Ying-Jie
Since the attribute aggregation of the evidential reasoning approach with multiple-level hierarchical structures is implemented in a recursive way, the aggregation times are the number of the branch nodes of attribute tree, which results in a large amount of calculation. To reduce the amount of calculation, a non-recursive aggregation approach is proposed and its time complexity is compared with that of the recursive approach. To explore the accuracy and the nonlinear characteristics of the non-recursive approach to dealing with different belief distributions, the belief degrees and the utilities of the aggregated attribute are calculated in terms of the belief structures of harmony, quasi-harmony and conflict, respectively. The nonlinear characteristics of the two aggregation approaches are analyzed based on the comparison of formula derivation and experimental results, and the accuracy of the non-recursive approach is examined from the angle of relative errors. The experimental results and the numeric example show the effectiveness of the proposed approach.
2014 Vol. 27 (4): 313-326 [Abstract] ( 544 ) [HTML 1KB] [ PDF 923KB] ( 590 )
Surveys and Reviews
327 A Survey of Machine Learning Algorithms for Big Data
HE Qing, LI Ning, LUO Wen-Juan, SHI Zhong-Zhi
With the explosive growth of the industry data, more and more attention is paid to big data. However, due to the volume, complex and fast-changing characteristics of big data, traditional machine learning algorithms for small data are not applicable. Therefore, developing machine learning algorithms for big data is a research focus. In this paper, the state-of-the-art machine learning techniques for big data are introduced and analyzed. As parallelism is a mainstream strategy for applying machine learning algorithms to big data, some parallelism strategies are described in detail as well. Finally, the challenges of applying machine learning to big data and some interesting research trends of machine learning in big data are pointed out.
2014 Vol. 27 (4): 327-336 [Abstract] ( 1170 ) [HTML 1KB] [ PDF 525KB] ( 5560 )
Researches and Applications
337 Dynamic Quotient Topology Model and Its Application to Optimal Path Finding
QI Ping, LI Long-Shu
To settle the problem solving under dynamic conditions, according to the fact that the topological structure changes with time, the traditional theory of quotient space is extended by using the trust model in sociology for reference. Based on the creditability evaluation of nodes by Bayesian model, a kind of dynamic quotient topology model based on the trust mechanism is proposed, and then this model is applied to optimal path finding. Theoretical analysis and simulation results prove that the proposed model can efficiently enhance the path reliability and meet the requirement of dynamic problem solving with fewer time costs.
2014 Vol. 27 (4): 337-344 [Abstract] ( 390 ) [HTML 1KB] [ PDF 510KB] ( 694 )
345 Image Classification Method by Combining Multi-features and Sparse Coding
LUO Hui-Lan, GUO Min-Jie, KONG Fan-Sheng
Using a single image feature to describe the image content is one-sided because of the insufficient information. Besides, the single coding method usually loses the spatial information. To solve these problems, an approach of integrating multi-features and sparse coding methods is proposed. Images are firstly divided into sub regions according to the spatial pyramid, and then the complementary advantages of scale invariant feature transform and the histogram of oriented gradients features are combined to produce various feature sets. Then, different clustering methods are used on different feature sets to acquire different codebooks. Next, two sparse coding methods, locality constrained linear coding and sparse coding based on each codebook are further employed respectively to get various image description sets. Finally, linear support vector machines are applied to image classification, and a voting method is used to determine the final classification. Experimental results show that the proposed method has good accuracy and robustness compared with some state-of-the-art methods.
2014 Vol. 27 (4): 345-355 [Abstract] ( 572 ) [HTML 1KB] [ PDF 1431KB] ( 846 )
356 A Construction Method of Fault-Tolerant Topology for Multi-agent Systems
WANG Qiang, CHEN Jie, FANG Hao
The fault-tolerant topology is important for multi-agent systems to complete the established objectives and tasks based on distributed cooperative control. The fault-tolerant topology of multi-agent systems is studied. A clustering algorithm based on weight is presented, and the hierarchical topology is constructed by distributed extraction of the virtual backbone network. A bi-connected network is realized using a distributed algorithm which combines articulation node-based algorithm with control of relay nodes on the basis of hierarchical structure. Application of the method to the networked fire-control system is represented. Experimental results show that the proposed methods enhance the fault tolerance of topology for multi-agent systems, which lays the foundations for the networked unmanned warfare theory into practice.
2014 Vol. 27 (4): 356-362 [Abstract] ( 463 ) [HTML 1KB] [ PDF 505KB] ( 959 )
363 Correlation Space Embedding Algorithm and Its Application to Image Retrieval
ZHUANG Ling, WANG Chao, ZHOU Feng, LU Wei-Ming, WU Jiang-Qin
An effective approach to semantic-based image retrieval is to find the correlation between low-level visual features and high-level semantics expressed by free text. Inspired by kernel method and graph Laplacian, the correlation space embedding algorithm(CSEA) is proposed in this paper. The latent semantic indexing and the visual word are used to construct the correlation between low-level image feature and semantic text feature which are heterogeneous with each other. The underlying cross-modal relationship between the free text and the image is established,and then the semantic-based image retrieval can be realized naturally. The consistency of manifold structure is regarded as a prior constraint in CSEA. By using CSEA, both the low-level image feature and the semantic text feature are embedded into a same intermediate space. Compared with the canonical correlation analysis, the proposed method models the correlation between two different feature spaces and preserves the manifold structure of each data distribution. Thus, the reliability of the proposed algorithm is improved. The experimental results show the effectiveness and the feasibility of the proposed algorithm in image retrieval.
2014 Vol. 27 (4): 363-371 [Abstract] ( 377 ) [HTML 1KB] [ PDF 1258KB] ( 670 )
372 Dynamic Granular Support Vector Machine Learning Algorithm
CHENG Feng-Wei, WANG Wen-Jian , GUO Hu-Sheng
Granular support vector machine (GSVM) is effective when dealing with distribution uniform datasets. However, the distribution of the dataset in the real world is unpredictable, and the density is uneven. In this paper, a dynamic granular support vector machine learning algorithm(DGSVM) is proposed. According to the different distribution of the granules, some granules are divided automatically and SVM training is performed on different levels of granule space. The experimental results on benchmark datasets demonstrate that DGSVM algorithm obtains better classification performance compared with GSVM.
2014 Vol. 27 (4): 372-377 [Abstract] ( 458 ) [HTML 1KB] [ PDF 449KB] ( 969 )
378 Weighted Multiscale and Multiresolution Face Description and Recognition Method
ZENG Zhi-Yong, LIU Shi-Gang

To overcome the effect of different illuminations and expressions on the recognition results of face images, a weighted multiscale and multiresolution face description and recognition method is presented. Multiresolution analysis is firstly employed to decompose a image into subimages, and three low frequence subbands with different scales are selected to construct multi-scale and multi-resolution image sequences. Then, the sign components of gray level difference between central pixel and its neighbors for each image in image sequences are encoded to express the importance of local face structure. Next, the magnitude components of gray level difference between central pixel and its neighbors are used as the weight of local binary pattern. Finally, block-based fisher linear discriminant analysis is utilized to reduce dimensions of the feature descriptor and enhance its discriminative ability. Experimental results on ORL and FERET face databases show that the proposed method gets significant performance improvement.
2014 Vol. 27 (4): 378-382 [Abstract] ( 417 ) [HTML 1KB] [ PDF 440KB] ( 757 )
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