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....

2017 Vol.30 Issue.2, Published 2017-02-28 Orignal Article        Orignal Article 97 Image Denoising Algorithm Based on Composite Convolutional Neural Network Lü Yongbiao, ZHAO Jianwei, CAO Feilong According to the theory of deep learning, the process of image denoising can be regarded as a fitting process of a neural network. In this paper, an image denoising algorithm based on composite convolutional neural network is proposed through constructing a simple and efficient composite convolutional neural network. The first stage includes two convolutional neural networks with two layers. Some initial convolutional kernels of convolutional neural network with three layers in the second stage are trained by these two networks, respectively. The training time in the second stage is decreased and the robustness of the network is enhanced. Finally, the learned convolutional neural network in the second stage is applied to denoise a new image with noises. Experimental results show that the proposed algorithm is comparable to state of the art image denoising algorithms in peak signal to noise ratio(PNSR), structure similarity, and root mean square error(RMSE). Especially, when the noises get heavier, the proposed algorithm performs better with less training time. 2017 Vol. 30 (2): 97-105 [Abstract] ( 1380 ) [HTML 1KB] [ PDF 2206KB] ( 1446 ) 106 Low Rank Projection Least Square Regression Subspace Segmentation for Gene Expression Data CHEN Xiaoyun, XIAO Bingsen, LIN Liyuan The traditional clustering methods are inefficient due to high dimension and redundancy, small sample size and noise of the gene expression data. Subspace segmentation is an effective method for high dimensional data clustering. However, the performance of clustering is reduced by using subspace segmentation on the gene expression data directly. To cluster the gene expression data more effectively, low rank projection least square regression subspace segmentation method(LPLSR) is proposed. The improved low rank method is utilized to project gene expression data into the latent subspace to remove the possible corruptions in data and get a relatively clean data dictionary. Then, least square regression method is employed to obtain the low-dimension representation for data vectors and the affinity matrix is constructed to cluster the gene data. The experimental results on six public gene expression datasets show the validity of the proposed method. 2017 Vol. 30 (2): 106-116 [Abstract] ( 662 ) [HTML 1KB] [ PDF 977KB] ( 621 ) 117 Methods for Personalized Drug Effectiveness Prediction in Cancer Precision Medicine WANG Hongqiang, GU Kangshen How to efficiently predict individual drug effectiveness is a hot topic of cancer research. In this paper, the way to precisely estimate and predict the effectiveness of targeted-drugs of a specific patient is analyzed by reviewing and examining the basic theory and the development of targeted treatment of cancer. A paradigm of genetic testing, cancerization genetic testing mode, is proposed. Recently developed high-throughput sequencing technique is utilized and molecular pathology of individual patient based on artificial intelligence are systematically analyzed and precisely diagnosed in this mode. The conventional genetic testing paradigm only focuses on mutation detection using low-throughput sequencing technologies with one-sided and seriously biased diagnosis and the shortcoming is overcome by the cancerization testing model with over-simplicity for realizing the precision medicine. Furthermore the clinical practice of the proposed mode is discussed. 2017 Vol. 30 (2): 117-126 [Abstract] ( 640 ) [HTML 1KB] [ PDF 1479KB] ( 1416 ) 127 Deep Hierarchical Feature Extraction Algorithm LI Qin, YOU Xiong, LI Ke, TANG Fen Extracting the image features with strong representation is critical to complete different vision tasks. The traditional features only describe one aspect of the image information, and therefore their representation capability is limited. In this paper, deep hierarchical feature(DHF) extraction algorithm based on the convolutional neural networks(CNN) is proposed. The essential information hidden inside the image is effectively mined by abstractly expressing the image in different layers. Firstly, the feature maps of the image are created based on CNN, and those in the convolutional layers are selected to construct the hierarchical structure of the image. Then, the best layer combination is determined according to the matching experiment. The feature maps in the low layers are described by the information entropy, and the ones in high layers are described by averaging the pixels in specified region. the DHF with strong representative capability is ultimately constructed. The experiment demonstrates that the proposed DHF has evident advantages compared with the existing features, and it can complete the matching task with high efficiency. 2017 Vol. 30 (2): 127-136 [Abstract] ( 993 ) [HTML 1KB] [ PDF 2011KB] ( 2194 ) 137 A Survey on Axiomatic Characterizations of Rough Approximation Operators WU Weizhi Lower and upper approximation operators are the foundation in the study of theoretic aspect of rough set theory as well as its practical applications. One of the main directions of the theoretic study of rough sets is the axiomatic characterization of rough approximation operators. Based on various binary relations, constructive definitions of classical rough approximation operators, rough fuzzy approximation operators, and fuzzy rough approximation operators are firstly introduced. Axiomatic characterizations of these approximation operators are then summarized and analyzed. Finally, perspectives and comparison of rough set approximation operators with other mathematical structures are discussed. 2017 Vol. 30 (2): 137-151 [Abstract] ( 672 ) [HTML 1KB] [ PDF 578KB] ( 405 ) 152 Route Planning Method for Unmanned Aerial Vehicle Based on Cultural Algorithm LI Ming, JIANG Leqi, CHEN Hao The existing route planning methods can not meet the optimal path and real-time requirements simultaneously. A method based on cultural algorithm is proposed to solve the problem of unmanned aerial vehicle(UAV) online path planning. According to the characteristics of cultural algorithm, online route planning method is combined with offline route planning method and they are fused into the population space of cultural algorithm. By extracting the knowledge, the situation knowledge is extracted from the initial path information, and the normative knowledge is recovered from the variation ranges of nodes. The planning space is limited by the knowledge and the time of planning is reduced. Different methods are combined to remedy the deficiencies of the existing methods. The experiment shows the proposed method searches target effectively in complex dynamic environments, and its planning speed is higher than that of other online route planning algorithms. Moreover, it can satisfy the real-time requirement, plan the shorter path and shorten the aircraft mission time. 2017 Vol. 30 (2): 152-161 [Abstract] ( 574 ) [HTML 1KB] [ PDF 1495KB] ( 473 ) 162 Fast Algorithm for Computing Approximations in Dominance-Based Rough Set WANG Shu, LI Tianrui The classical rough set can not process preference-ordered data. Dominance-based rough set (DRST) overcomes this drawback. The data processing efficiency can be improved by reducing the time of computing approximations. A fast algorithm for computing approximations is presented. The approximations are acquired quickly while objects and attributes being added simultaneously in DRST. The definitions of parameters related to approximations are revised in the proposed fast algorithm and thus approximations can be calculated by parameters as few as possible. Consequently, the calculation is simplified and accelerated, and the memory consumption is reduced as well. The experimental results demonstrate that the proposed algorithm is faster than other algorithms and it is especially efficient with larger data sizeand data label. 2017 Vol. 30 (2): 162-170 [Abstract] ( 507 ) [HTML 1KB] [ PDF 740KB] ( 376 ) 171 Parallel Ensemble Learning Algorithm Based on Improved Binary Glowworm Swarm Optimization Algorithm and BP Neural Network LI Jingming, NI Zhiwei, ZHU Xuhui , XU Ying The traditional back propagation(BP) neural network has low learning speed and calculution accuracy and it is easy to fall into local solution. Aiming at these defects, a parallel ensemble learning algorithm based on improved binary glowworm swarm optimization algorithm(IBGSO) and BP neural network is proposed. Firstly, a kind of improved binary glowworm swarm algorithm is constructed based on Gauss variation function as probability mapping function, and the validity of the algorithm is analyzed theoretically. Secondly, The IBGSO algorithm and BP neural network are combined to construct a parallel ensemble learning algorithm. Finally, the parallel ensemble learning algorithm is applied to the assessment of agricultural drought disaster. The experimental results show that the algorithm has advantages over the traditional algorithms in terms of convergence speed and operation accuracy. Therefore, IBGSO-BP algorithm can effectively improve the accuracy of agricultural drought assessment. 2017 Vol. 30 (2): 171-182 [Abstract] ( 622 ) [HTML 1KB] [ PDF 891KB] ( 700 ) 183 Path Coverage Scheme Based on Fuzzy Particle Swarm Optimization Algorithm for Directional Sensor Networks ZHANG Juwei, WANG Yu, YANG Ting By utilizing fuzzy data fusion rules, a fuzzy perception model for directional sensor nodes is built to reduce the network uncertain region. Aiming at path coverage problems of directional sensor networks, a path coverage enhancement algorithm of the directional sensor networks based on fuzzy particle swarm optimization is proposed. The formed n-dimension problem is transformed to one-dimension problem to improve the coverage area of single sensor node and thereby increase the network coverage. For the directional sensor network nodes of adjustable perception direction, the simulation experiment is carried out by comparing the proposed algorithm with the existing algorithms under random deployment. The results show that the proposed algorithm can effectively improve the path coverage of the directional sensor networks, have a faster convergence rate and prolong the network life time. 2017 Vol. 30 (2): 183-192 [Abstract] ( 500 ) [HTML 1KB] [ PDF 2026KB] ( 585 )

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