Started in January,1974(Monthly)
Supervised and Sponsored by Chongqing Southwest Information Co., Ltd.
ISSN 1002-137X
CN 50-1075/TP
Current Issue
Volume 43 Issue 2, 01 December 2018
Research and Advances on Deep Learning
SUN Zhi-yuan, LU Cheng-xiang, SHI Zhong-zhi and MA Gang
Computer Science. 2016, 43 (2): 1-8.  doi:10.11896/j.issn.1002-137X.2016.02.001
Abstract PDF(951KB) ( 1300 )   
References | Related Articles | Metrics
Deep learning (DL) is a recently-developed field belonging to machine learning.It tries to mimic the human brain,which is capable of processing the complex input data fast,learning different knowledge intellectually,and solving different kinds of complicated human intelligence tasks well.Recently,with the advent of a fast learning algorithm for DL,the machine learning community set off a surge to study the theory and applications of DL since it has many advantages.Practice shows that deep learning is a kind of high efficient feature extraction method,which can detect more abstract characteristics and realize the essence of the data,and the model constructed by DL tends to have stronger genera-lization ability.Due to the advantages and wide applications of deep learning,this paper attempted to provide a started guide for novice.It presented a detailed instruction of the background and the theoretical principle of deep learning,its emblematic models,its representative learning algorithm,the latest progress and applications.Finally,some research directions of deep learning that are deserved to be further studied were discussed.
Prediction Algorithm for Seasonal Satellite Parameters Based on Time Series Decomposition
ZHOU Feng and PI De-chang
Computer Science. 2016, 43 (2): 9-12.  doi:10.11896/j.issn.1002-137X.2016.02.002
Abstract PDF(339KB) ( 524 )   
References | Related Articles | Metrics
The prediction of satellite seasonal parameters plays a key guiding role for the satellite fault prediction.Considering the low prediction precision problem of satellite seasonal parameters,a prediction method for seasonal satellite parameters based on time series decomposition was proposed.Firstly,the wavelet analysis method is used to eliminate noise and extract the cycle from the parameter sequence in the frequency domain.Then,with the time series decomposition method in the time domain,the trend item and random item are generated.Thus the gray model(GM) and the auto regressive moving average model(ARMA) are used to predict these items respectively according to their characteristics.Finally,all the prediction parts are combined and the final predictive value is got.The contrast experimental prediction and analysis of satellite remote sensing data verify the effectiveness of the proposed method.
Operational Architecture Modeling Based on Information Flow
YANG Ying-hui, LI Jian-hua, NAN Ming-li and TIAN Yan-tao
Computer Science. 2016, 43 (2): 13-18.  doi:10.11896/j.issn.1002-137X.2016.02.003
Abstract PDF(1002KB) ( 495 )   
References | Related Articles | Metrics
To depict information flow structure of systematic operations synthetically and describe complex dynamic information disposal and alternation situation among operational activities and units carefully,this paper introduced information flow analysis method.Three information flow factors of entity,relationship and attribute were defined,relationships among entities were analyzed and four information mapping rules of activity→activity,activity→node,node→activity and node→node were given with information flow as main line.According to basic modeling flows,models of ope-rational activity,node and information alternation were built.Finally,an example of aerial assault operation was taken for validation as simulation.The results show that the method and models proposed are feasible and effective,and can describe complex information alternation relationship in operational systems and present dynamic process of information flow impenetrating operational activities,which provide a new method for operational architecture modeling and information requirement analysis.
Research on Multi-source Heterogeneous Information Dynamic Integration Technology for Industrial-chain Coordination SaaS Platform
LV Rui, WANG Shu-ying, SUN Lin-fu and PAN Hua
Computer Science. 2016, 43 (2): 19-25.  doi:10.11896/j.issn.1002-137X.2016.02.004
Abstract PDF(1141KB) ( 415 )   
References | Related Articles | Metrics
In order to reduce the costs of time and space for exchanging large amounts of data for cooperative enterprise’s internal systems and industry chain collaboration SaaS platform,and improve the accuracy of real-time data query,the multi-source heteroge-neous information integration service technology was researched.Multi-source heterogeneous information integration model for industrial chain collaboration SaaS platform was established.Multi-source heterogeneous information conversion and registration algorithms were established.User’s identity and business driven dynamic invocation algorithm was also proposed.Multi-source heterogeneous real-time accurate information inquiry service of industry chain collaboration SaaS platform was realized.The applications of the model and algorithms in service provider’sparts inventory management of automotive parts industry chain collaboration platform prove the technology is feasible and effective.
Multi-pattern Matching Algorithm Based on Coding Association
ZHU Yong-qiang and QIN Zhi-guang
Computer Science. 2016, 43 (2): 26-30.  doi:10.11896/j.issn.1002-137X.2016.02.005
Abstract PDF(425KB) ( 414 )   
References | Related Articles | Metrics
Multi-pattern matching algorithms often use finite state automaton to implement parallel matching of multiple pattern strings.When multi-pattern matching algorithm based on finite state automaton is applied into the Chinese,it will lead to storage space expansion.Aiming to solve this problem,this improved algorithm constructs automatic state machine by using split coding of Chinese characters to save storage space and designs failure jump table based on coding association,and uses heuristic jumping rules to improve time performance of matching.Finally,compared to other algorithm,smaller space consumption and faster speed in Chinese environment of this improved algorithm were proved by simulation.
Comparison of Chinese Anaphora Resolution Models
ZHOU Xuan-yu, LIU Juan, LUO Fei, LIU Yang and YAN Han
Computer Science. 2016, 43 (2): 31-34.  doi:10.11896/j.issn.1002-137X.2016.02.006
Abstract PDF(426KB) ( 1008 )   
References | Related Articles | Metrics
The task of anaphora resolution is determining linguistic expressions that refer to the same real-world entity in a text.Compared to the amount of research on English anaphora resolution,relatively little work has been done on Chinese anaphora resolution.To our best knowleage,there is no fair comparison on the Chinese anaphora resolution models,mainly because of the useage of different data and features.We aimed to gain insights into the factors that affect the performance of the models and the advantages of each model via experiment with the same data and features in the ACE2005 Chinese corpus.
Fingerprint Focal Point Detection Method Based on Map of Cross Points’ Weights
GUO Xi-feng, ZHU En, ZHOU Si-hang, SHEN Xiao-long and YIN Jian-ping
Computer Science. 2016, 43 (2): 35-37.  doi:10.11896/j.issn.1002-137X.2016.02.007
Abstract PDF(602KB) ( 451 )   
References | Related Articles | Metrics
The reference point is essential for fingerprint registration and recognition since it can effectively reduce the identification time on automatic fingerprint identification system with large scale database.As an extremely stable refe-rence point,the fingerprint focal point still needs more accurate detection methods.In this paper,an accurate fingerprint focal point detection method based on map of cross points’ weights was proposed.First,the definition and characteristics of the proposed map of cross points’ weights are introduced.For the block-wise orientation field,from each block’scenter,a straight line perpendicular to the block’s direction is produced,and each pair of such straight lines may have a cross point.By giving each cross point a weight according to its contribution to the expected position of fingerprint focal point,the map of cross points’ weights is finally determined.The proposed map improves the accuracy and stability of focal point detection method since it can generally ensure the ideal focal point to be included in its maximum density region.Then an iteration scheme is utilized on map of cross points’ weights to locate the focal point precisely.Experimental results on FVC2000 Db2a fingerprint database validate that the proposed method outperforms the others in terms of accuracy with acceptable cost of computational time.
Super-resolution Mapping Using Pixel-swapping Based on Integration of Coarse-scale Spatial Heterogeneity and Fine-scale Spatial Homogeneity
HU Jian-long, LI De-yu and BAI He-xiang
Computer Science. 2016, 43 (2): 38-40.  doi:10.11896/j.issn.1002-137X.2016.02.008
Abstract PDF(769KB) ( 384 )   
References | Related Articles | Metrics
Spatial dependence characterization plays a key role for super-resolution mapping.Experiments and observations show that the coarse-scale spatial heterogeneity characterization can better describe spatital heterogeneity of ground objects between different classes,while the fine-scale spatial homogeneity characterization can better describe spatital homogeneity of ground objects in the same class.This paper proposed a new algorithm for super-resolution mapping using pixel-swapping strategy based on the combination of spatial heterogeneity at coarse scale and homogeneity at fine scale.The integration of coarse-scale heterogeneity and fine-scale homogeneity will better represent features of complex land cover.Experimental results on the sythentic image further validate the effectiveness of algorithm and it achieves higher precision under the premise of the same fraction information.
Data Compression with Attribute Homomorphism in Information Systems
HAO Yan-bin, GUO Xiao and YANG Nai-ding
Computer Science. 2016, 43 (2): 41-46.  doi:10.11896/j.issn.1002-137X.2016.02.009
Abstract PDF(492KB) ( 468 )   
References | Related Articles | Metrics
Data compression is an important topic in data processing and homomorphism is considered as an effective tool for data compression.This paper defined the attribute isomorphism and attribute homomorphism for information system based on the functional dependency relation over attributes.It then investigated major properties of the attribute homomorphism.By using the attribute equivalence to acquire the ideal homomorphism,it achieved lossless compression for information system.Finally,this paper provided a method to measure the ideal level of any attribute homomorphism by comparing distance between the original system and the image information system.
Hyperspectral Remote Sensing Image Data Processing Research and Realization Based on CPU/GPU Heterogeneous Model
TANG Yuan-yuan, ZHOU Hai-fang, FANG Min-quan and SHEN Xiao-long
Computer Science. 2016, 43 (2): 47-50.  doi:10.11896/j.issn.1002-137X.2016.02.010
Abstract PDF(440KB) ( 567 )   
References | Related Articles | Metrics
In recent years,the development of new high-performance heterogeneous computing based on GPU provides good oppotunities in many application areas.Domestic and foreign remote sensing experts have started to introduce it to solve the issues like computation intensive and difficult real-time processing caused by high-dimensional space features of hyperspectral image.In this brief introduction to hyperspectral remote sensing and CPU/GPU heterogeneous computing model,we summarized hyperspectral data processing status and problems based on CPU/GPU heterogeneous pattern in recent years,and for small desktop supercomputer with shared storage,realized parallelization of hyperspectral imaging MNF dimensionality reduction on CPU/GPU heterogeneous model,and verified the development potential of heterogeneous patterns in the field of hyperspectral remote sensing processing by contrasting with the sequential program and OpenMP.
Recognition of Prosodic Phrases Based on Unlabeled Corpus and “Adhesion” Culling Strategy
QIAN Yi-li and CAI Ying-ying
Computer Science. 2016, 43 (2): 51-56.  doi:10.11896/j.issn.1002-137X.2016.02.011
Abstract PDF(493KB) ( 410 )   
References | Related Articles | Metrics
Obtaining large-scale annotated corpus manually is very difficult and has some disadvantages.Based on the pause role of punctuation,this paper proposed a prosodic phrase recognition method which uses unlabeled corpus and “adhesion” culling strategy.In the method,punctuation is graded and given different weights when it is used to simulate the prosodic boundaries.For recognizing prosodic phrase boundaries automatically,a max entropy model is constructed based on an unlabeled corpus and a Top-K method is also used.According to the mutual information of two contiguous part of speech tagging,words are bundled into adhesion units and the prosodic boundaries appear in it are eliminated.The experimental results show that hierarchical use of punctuation and “adhesion” culling strategy can improve the performance of the model significantly.The method can obtain better recognition results.
Social Recommendation Combining Global and Dual Local Information
QIAN Fu-lan and LI Qi-long
Computer Science. 2016, 43 (2): 57-59.  doi:10.11896/j.issn.1002-137X.2016.02.012
Abstract PDF(316KB) ( 463 )   
References | Related Articles | Metrics
With the rapid growth of Web2.0,social recommendation has become one of the hot research topics in the last few years.It is the key point to improve recommender systems using social contextual information in a more efficient way.The existing social recommendation approaches mainly take advantage of user’s direct connection(explicit relation).This paper detailed social relation as explicit relation and implicit relation and obtained the user’s reputation by using his/her historic records.Then we proposed a recommendation framework capturing user’s global social relation(reputation)and local social relation(explicit relation and implicit relation).Using two real datasets,Douban and Epi-nions,we conducted a experimental study to investigate the performance of the proposed model GDLRec.We compared our approach with existing representative approaches.The results show that GDLRec outperforms other methods in terms of prediction accuracy.
Infrared Dim and Small Target Detection Based on Background Prediction by Wavelet Filter
JIAO Jiao, XIE Yong-jie, ZHANG Hua-liang and ZHANG Song
Computer Science. 2016, 43 (2): 60-63.  doi:10.11896/j.issn.1002-137X.2016.02.013
Abstract PDF(864KB) ( 392 )   
References | Related Articles | Metrics
Dim and small target detection is a critical technology in the infrared image field. The target can be detected effectively from the background,which is important for the work of target tracking and recognition.Aiming at defects of the existing target detection technology,an approach for infrared dim and small target detection based on background prediction by wavelet filter was proposed.The approach utilizes wavelet filter to remove the target as noise,then the background image and the foreground image are estimated,finally infrared dim and small target can be extracted effectively by connecter screening and contrast thresholding.The approach was tested on the real images of photoelectric theodolite.Experiments show effectiveness of the method on anti-noise,background prediction,and infrared dim and small target detection.
Capsule Defects Detection Based on Stacked Denoising Autoencoders
WANG Xian-bao, HE Wen-xiu, WANG Xin-gang, YAO Ming-hai and QIAN Yun-tao
Computer Science. 2016, 43 (2): 64-67.  doi:10.11896/j.issn.1002-137X.2016.02.014
Abstract PDF(604KB) ( 437 )   
References | Related Articles | Metrics
At present defects of capsules are detected mainly by manual operation,which is time-consuming and needs high labor costs,besides,it is easily misled by subjective factors.This paper proposed a method of detection of capsules surface defects based on stacked denoising autoencoders (SDAE).Our method firstly establishes deep autoencoders networks and trains using a denoising criterion according to the defect samples to obtain the initial weights at first.Then,BP algorithm fine-tunes the network parameters to get the mapping relationship between the training sample and defect-free template.Finally, defect detection of the testing samples is finished by comparing the reconstruction image and defect image.Experimental results show that SDAE perfectly establishes the mapping relationship,which is robust and stable to noise,and can quickly detect defects with high accuracy.
Rough Set Algebra of Multi-granulation
KONG Qing-zhao and WEI Zeng-xin
Computer Science. 2016, 43 (2): 68-71.  doi:10.11896/j.issn.1002-137X.2016.02.015
Abstract PDF(250KB) ( 443 )   
References | Related Articles | Metrics
It is well known that a rough set algebra is a set algebra with added dual pair of rough approximation operators.On the one hand,we discussed the classical rough set algebra of multi-granulation by axiomatic approach.It is shown that the classical rough set algebra does not possess good properties.On the other hand,we defined the concept of monotone equivalence relations.Moreover,multi-granulation approximation operators based on monotone equivalence relations were defined.We discussed the properties of the rough set algebra based on monotone equivalence relations and got many excellent results.
Improving Recommendation Diversity via Probabilistic Selection
ZHANG Dong, CAI Guo-yong and XIA Bin-bin
Computer Science. 2016, 43 (2): 72-77.  doi:10.11896/j.issn.1002-137X.2016.02.016
Abstract PDF(503KB) ( 434 )   
References | Related Articles | Metrics
Typical recommendation algorithms focus on optimizing the accuracy of recommendation lists,however,diversity is also considered as a key property to measure the quality of recommendation lists from both user and system perspective.Many list diversification techniques improve diversity by re-ranking items.In this paper,a new probabilistic selection model for improving the diversity of recommendation lists was proposed.This model transfers the list generation process to N-times probabilistic selection process,and each selection includes two steps:genre selection and item selection.For the genre selection phase,genre information of items is included to compute user-genre probabilistic matrix,and a genre is chosen based on this matrix.For the item selection phase,three properties including estimated score of items,historical popularity of items,and recommending popularity of items are considered for item re-scoring.The item with the highest re-computed score will be selected into the recommendation list.The trade-off between diversity and accuracy can be controlled by changing threshold value TR.Experiments on two movie recommendation datasets show that our model can effectively improve recommendation lists diversity.At the same time,the comparative experiments show that our model outperforms re-ranking method in almost all experimental results,except the case of individual diversity for matrix factorization.
Microblog Topic Evolution Algorithm Based on Retweeting Relationship
XU Wei, ZHAO Bin and JI Gen-lin
Computer Science. 2016, 43 (2): 78-82.  doi:10.11896/j.issn.1002-137X.2016.02.017
Abstract PDF(528KB) ( 599 )   
References | Related Articles | Metrics
Existing work has been focused on topic evolution of long text.This paper aimed to that of short text.We proposed a microblog topic evolution algorithm MTERR based on retweeting relationship.Firstly,we utilized a topic model to obtain topic information from microblog messages by combining retweeting features and time characteristics.Then,we built a topic correlation function to generate a topic evolution topological graph by incorporating topic content and retweeting relationship.Experiments on the real-world microblog datasets show the feasibility and effectiveness of our proposed method.
Granularity Reduction of Variable Precision Pessimistic Multi-granulation Rough Set Based on Granularity Entropy of Lower Approximate Distribution
MENG Hui-li, MA Yuan-yuan and XU Jiu-cheng
Computer Science. 2016, 43 (2): 83-85.  doi:10.11896/j.issn.1002-137X.2016.02.018
Abstract PDF(321KB) ( 396 )   
References | Related Articles | Metrics
The lower approximate distribution reduction was introduced into the variable precision pessimistic multi-granulation rough set.The granularity entropy of the lower approximate distribution of the variable precision pessimistic multi-granulation rough set was defined.The importance of a granularity was also defined based on the granularity entropy of the lower approximate distribution,and a heuristic granularity reduction algorithm of variable precision pessimistic multi-granulation rough set was presented.The experimental results show the validity of the algorithm.
Improved WPR Algorithm Based on Referenced Frequency in Recent Search Cycle
WANG Xu-yang and REN Guo-sheng
Computer Science. 2016, 43 (2): 86-88.  doi:10.11896/j.issn.1002-137X.2016.02.019
Abstract PDF(252KB) ( 400 )   
References | Related Articles | Metrics
For the topic drift and bias towards the old pages of WPR(Weighted PageRank) algorithm exist in the Web search,consolidated two factors of Web pages’ topic features and referenced frequency in recent search cycle,we proposed an improved algorithm WTFPR(Weighted Topic Frequency PageRank).The algorithm uses improved TD-IDF algorithm to solve relevance of page by content analysis to reduce the topic drift. The algorithm improves the PR value of new and has high quality by referenced frequency of pages in recent search cycle,reducing bias towards the old pages.Simulation results show that the improved algorithm obtaines better results compared to WPR.
Hyperspectral Image Classification Method Based on Watershed Segmentation and Sparse Representation
SHU Su and YANG Ming
Computer Science. 2016, 43 (2): 89-94.  doi:10.11896/j.issn.1002-137X.2016.02.020
Abstract PDF(1022KB) ( 419 )   
References | Related Articles | Metrics
In recent years,the classification has attracted wide attention.Many machine learning methods have been applied in hyperspectral image classification,such as SVM,neural network and decision tree.But in the hyperspectral image,different materials may have the same spectra and the same material in different locations may have different spectra,consequently bringing a challenge for accurate classification of hyperspectral image.So,we made use of the spatial information extracted from the watershed segementation and the sparse representation to get a more accurate classification results.Firstly,we extracted regional information from hyperspectral image by watershed segementation,then classificated all the samples in a region once.The effectiveness of our proposed method was evaluated via two images.And the results show that it exhibits state-of-the-art performance.
Attribute Granulation Based on Attribute Discernibility and AP Clustering
ZHU Hong and DING Shi-fei
Computer Science. 2016, 43 (2): 95-97.  doi:10.11896/j.issn.1002-137X.2016.02.021
Abstract PDF(314KB) ( 418 )   
References | Related Articles | Metrics
This paper put forward a kind of attribute granulation method based on attribute discernibility and AP clustering.The method calculates the similarity of attributes according to attribute discernibility firstly,and then clusters attributes into several groups through affinity propagation clustering algorithm.At last,representative attributes are produced through some algorithms to form a coarser attribute granularity.The method is more efficient than traditional attribute reduction algorithm for large data set.It has obvious advantages under the condition of less strict precision of attribute granularity.
Homotopy Support Vector Machine
JIANG Jin-chao and ZHANG Rui
Computer Science. 2016, 43 (2): 98-100.  doi:10.11896/j.issn.1002-137X.2016.02.022
Abstract PDF(226KB) ( 370 )   
References | Related Articles | Metrics
As a novel artificial intelligence technology,support vector machines (SVM) has been more and more widely used in various fields.Homotopy regularization method is a regularization method emerging in recent years,which has been widely used in the inverse problem.This paper applied the ideas of the homotopy regularization to the support vector machine,and established a new SVM.Meanwhile,we modified the most used Gaussian kernel.Compared with conventional regularization method,the biggest advantage of the new model is that the regularization parameter values range changes from infinite interval to a limited range (0,1),thus it greatly shortens the time of regularization parameter optimization.
Algorithm to Determine Number of Clusters for Mixed Data Based on Prior Information
PANG Tian-jie and ZHAO Xing-wang
Computer Science. 2016, 43 (2): 101-104.  doi:10.11896/j.issn.1002-137X.2016.02.023
Abstract PDF(301KB) ( 364 )   
References | Related Articles | Metrics
In cluster analysis,one of the most challenging and difficult problem is the determination of the number of clusters.The strategies for choosing initial prototypes randomly are used to determine the number of clusters in most of the existing methods,resulting in weak stability of iterations in clustering process.So we proposed an prior information based algorithm to determine the number of cluster for mixed data by using priori information which includes class labels to optimize initial prototype .Experiments show that the algorithm is effective.
Face Enhancement Algorithm with Variable Illumination Based on Improved Retinex
DU Ming and ZHAO Xiang-jun
Computer Science. 2016, 43 (2): 105-108.  doi:10.11896/j.issn.1002-137X.2016.02.024
Abstract PDF(1204KB) ( 650 )   
References | Related Articles | Metrics
In order to improve the overall effect of face images with variable illumination,this paper proposed a novel face enhancement algorithm with variable illumination based on single scale Retinex.Firstly,the face images are logarithmically transformed,and the image is transformed into frequency and low frequency part by curvelet transform.Se-condly,the bilateral filtering is used to denoise the high frequency while Kimmel variation model is used to smooth filtering low frequency part.Finally,the image is reconstructed,and Gamma is used to correct the image.The experimental results on Yale B database show that the proposed algorithm can prevent the “halos” phenomenon,and can restore the original face image,so the face image is more suitable for human eye observation.
Distributed Localization Scheme Based on Virtual Force in Wireless Sensor Networks
XIONG Zhe, JIA Jie and CHEN Jian
Computer Science. 2016, 43 (2): 109-112.  doi:10.11896/j.issn.1002-137X.2016.02.025
Abstract PDF(334KB) ( 412 )   
References | Related Articles | Metrics
Localization is a basic problem in wireless sensor networks.Traditional localization algorithms are usually based on centralized computing,which may result in high cost.For this reason,the least square method was first applied to estimate the preliminary position.Further,the localization model for all nodes based on virtual force was presented,and a virtual force based localization algorithm was proposed.By exchanging information between neighbor nodes in a distributed manner,our algorithm can effectively save localization communication cost.Finally,an update mechanism was proposed to elevate unknown node as anchor node,thus to accelerate localization process.Extensive simulations were presented to demonstrate the effectiveness of our distributed iterative localization algorithm.
xMAS-based Formal Verification of SpaceWire Credit Logic
LI Yan-chun, LI Xiao-juan, GUAN Yong, WANG Rui, ZHANG Jie and WEI Hong-xing
Computer Science. 2016, 43 (2): 113-117.  doi:10.11896/j.issn.1002-137X.2016.02.026
Abstract PDF(755KB) ( 353 )   
References | Related Articles | Metrics
SpaceWire protocol is a high-speed communication bus protocol applied to aerospace,so it is very important for communication system to ensure the reliability of the design.Due to the presence of a large number of queues,distributed control and concurrency,the traditional verification methods have incomplete defects and state explosion when model checking occurs.This paper presented a formal verification method of credit logic in SpaceWire communication system with xMAS model.xMAS model retains the structural information in lower level and can verify high-level attri-butes.The paper built an abstract xMAS model for credit logic and listed three key properties including sending,receiving and data consistency.Correctness of the properties was verified automatically by the ACL2 theorem proving tool.It can provide effective reference for system design under the guidance of verification.
Projection Based Algorithm for Link Prediction in Bipartite Network
GAO Man, CHEN Ling and XU Yong-cheng
Computer Science. 2016, 43 (2): 118-123.  doi:10.11896/j.issn.1002-137X.2016.02.027
Abstract PDF(823KB) ( 427 )   
References | Related Articles | Metrics
An algorithm for link prediction in a bipartite network was presented.In the algorithm we first mapped the bipartite network to unipartite one called projected graph.Based on the projected graph,we defined the concept of potential link.We performed the link prediction only within the potential links so as to reduce the computation time.We also defined the pattern covered by the potential links and the weight of the patterns.By calculating the weight of the patterns a potential link covers,the confidence of the potential link can be obtained,which can be used as the final score of link prediction.Experimental results show that our algorithm can get faster speed and higher quality of link prediction results.
Update Strategy of Communities Information Based on Variable Half-lifes in Opportunistic Networks
YAN Yu-dao and LIU Lin-feng
Computer Science. 2016, 43 (2): 124-128.  doi:10.11896/j.issn.1002-137X.2016.02.028
Abstract PDF(433KB) ( 385 )   
References | Related Articles | Metrics
Due to the incomplete use of historical information of nodes’ movements,to introduce the concept of group,every single person can be put into the connect of the community.Theoretically,the movement of nodes can be abstrac-ted as movement from one group to another.We proposed a new method mass-group detected with interest value (MDIR) to compute the utility value in opportunistic routing in social network.With the concept of interest value,dynamic communities can be detected precisely.Experiments indicate that in different intensiveness,MDIR performes better stability and efficiency compared with Epidemic,BDCR and SREP.
Degrees of Freedom of Coexistence Network with Point-to-Point and Two-user Broadcast Channels
LIU Feng, WANG Yuan-yuan and ZENG Lian-sun
Computer Science. 2016, 43 (2): 129-134.  doi:10.11896/j.issn.1002-137X.2016.02.029
Abstract PDF(446KB) ( 432 )   
References | Related Articles | Metrics
The degrees of freedom (DoF) of two networks with coexistence was investigated in this paper.One is a point-to-point(PTP) channel and the other is a two-user broadcast channel (BC).Four system models were considered.First,BC as se-condary network is not cognitive of PTP (primary network) transmitter’ message.Second,PTP as se-condary network is not cognitive of BC transmitters’ messages.Third,BC as secondary network is cognitive of PTP transmitter’ message.Fourth,PTP as secondary network is cognitive of BC transmitters’ messages.The achievable DoF was obtained by using interference alignment and interference neutralization.Besides,the outer bound was proven.Compared with the other three system models,the achievable bound is biggest when BC as secondary network is cognitive of PTP transmitters’ message.
Rumor Spreading Model Considering Conformity Phenomena in Complex Social Networks
ZHU Guan-hua, JIANG Guo-ping and XIA Ling-ling
Computer Science. 2016, 43 (2): 135-139.  doi:10.11896/j.issn.1002-137X.2016.02.030
Abstract PDF(509KB) ( 708 )   
References | Related Articles | Metrics
With the prevalence of the social networking services (SNS),the study of rumor spreading mechanisms and models for controlling the propagation of rumors is more and more important in human lives.Inspired by the conformity phenomena in the social networks,we established a new susceptible-infected-removed (SIR) rumor spreading model with consideration of the influences of the global information contained in the SNS’ message,the scale of SNS and the diversity of individuals.The simulation results show that the new proposed model can give a good description of the amplitude effect caused by the conformity phenomena in the rumor spreading process.
Fault Tolerant Regression Model for Sensing Data
ZUO Xiang-dong, WANG Kun and QIU Hui
Computer Science. 2016, 43 (2): 140-143.  doi:10.11896/j.issn.1002-137X.2016.02.031
Abstract PDF(335KB) ( 353 )   
References | Related Articles | Metrics
Sensors are always used to monitor external environment,whenever faults occur in sensors,there will be prediction errors.In order to improve the ability of fault tolerance when faults occur,this paper proposed a fault tolerant regression model for sensing data.Firstly,we analyzed linear models including least squares and ridge regression,and also analyzed statistics of regression model.Secondly,we analyzed the related statistics when sensors failed,and analyzed the lower and upper bounds of covariate matrix based on these statistics.Finally,we defined fault index based on covariate matrix,and transformed the optimum of model into minimize fault index and mean square error simultaneously.The experiments show that the proposed fault tolerant model has lower error than traditional least squares and ridge regression,and thus has better robustness when sensors fail.
Low Delay Topology Control Algorithm Based on Delivery Ratio Constraint in Wireless Sensor Networks
KONG Shan-shan, LIU Lin-feng and CHEN Hang
Computer Science. 2016, 43 (2): 144-147.  doi:10.11896/j.issn.1002-137X.2016.02.032
Abstract PDF(442KB) ( 325 )   
References | Related Articles | Metrics
In the paper,the objective and requirement of topology control were analyzed based on the application scenes of emergency data collection (such as earthquakes,fire alarms),then network model was constructed,and it was described formally and analyzed by mathematical theory.A low delay topology control algorithm based on delivery ratio constraint (LDBDC) was proposed accordingly.The algorithm can calculate the approximate optimal average number of hops for given area so as to obtain the length of virtual grid based on pre-given delivery ratio constraint.The simulation experiments suggest that LDBDC can obtain approximate optimal topology structure and obtain the minimum average delay satisfying the premise of delivery ratio constraint.
Identity Authentication Method Based on User’s Mouse Behavior
XU Jian, LI Ming-jie, ZHOU Fu-cai and XUE Rui
Computer Science. 2016, 43 (2): 148-154.  doi:10.11896/j.issn.1002-137X.2016.02.033
Abstract PDF(557KB) ( 764 )   
References | Related Articles | Metrics
Aiming at the problems of existing solution for identity authentication,the paper provided a method about the identity authentication based on mouse behaviour.First,it gave the authentication model based on the mouse action and the entity set,and then defined the mouse action by hierarchical division method.At the same time,it gave the definition of features related to the mouse action and the computing method.The paper used random forest classifier as the classification tool to solve the problem of data over-fitting and noise in the existing solution.In the phase of authentication,it used hierarchical classification-decision model for identity authentication.Finally,the paper analyzed the method through experiment,showing a better false rejection rate and false acceptance rate.
Dynamic Symbolic Taint Analysis of Binary Programs
ZHU Zheng-xin, ZENG Fan-ping and HUANG Xin-yi
Computer Science. 2016, 43 (2): 155-158.  doi:10.11896/j.issn.1002-137X.2016.02.034
Abstract PDF(425KB) ( 818 )   
References | Related Articles | Metrics
The dynamic taint analysis (DTA for short) technique is usually applied to track information flow and detect security vulnerabilities.It detects the vulnerabilities of program triggered by some test cases dynamically.Though its false positive rate is very low,its false negative rate is very high.Concerning this issue,the dynamic symbolic taint ana-lysis (DSTA for short) is an enhancement to dynamic symbolic analysis,which symbolizes the taint analysis to reduce false negative rate.The technique collects taint information according to taint propagating based on instructs,and makes symbolic risk rule to find some potential vulnerabilities by detecting whether the taint information breaks some risk rules.The experimental results show that this method not only ensures the advantage of DTA’s low false positive rate,but also reduces the disadvantage of DTA’s high false negative rate.The information of vulnerabilities,risks and taint data can be applied to generate test cases,which improves the test efficiency and reduces the redundancy of test case.
Anonymous Multi-user Searchable Encryption Scheme with Hierarchial Permission Management
DUAN Yang-yang and LI Shi-yang
Computer Science. 2016, 43 (2): 159-162.  doi:10.11896/j.issn.1002-137X.2016.02.035
Abstract PDF(446KB) ( 384 )   
References | Related Articles | Metrics
To deal with the existing problems of multi-user searchable encryption schemes in the cloud computing,we firstly proposed a basic multi-user searchable encryption scheme,and then extended this basic scheme to a multi-user searchable encryption scheme with anonymous hierarchical permission management.Compared with existing schemes,our scheme not only achieves privacy preserving in both of the searching content and users’ identities,but also allows the data owner to directly control the dynamic updating of query permission.Additionally,our scheme realizes the hie-rarchical users’ query permission management by adopting a specific query key generation rule.The security of our scheme was illustrated by security analysis.Performance evaluation and experimental result show that our scheme is practical and feasible.
XACML Policy Optimization Method Based on Redundancy Elimination and Attribute Numericalization
QI Yong, CHEN Jun and LI Qian-mu
Computer Science. 2016, 43 (2): 163-168.  doi:10.11896/j.issn.1002-137X.2016.02.036
Abstract PDF(553KB) ( 667 )   
References | Related Articles | Metrics
XACML (eXtensible Access Control Markup Language) has become one of main access control standards.Access control systems need effective XACML evaluation engine to ensure system availability.To solve the problem above,this paper optimized XACML policy from two aspects:redundancy elimination and attribute numericalization,based on the potential shortcomings of XACML itself.Redundancy elimination removes the redundant rules in the policies and the redundant states between the rules by applying rule compression method.Attribute numericalization transforms textuary attributes of XACML policies into numerical attributes,to make evaluation engine use effective numerical match,instead of inefficient string match.In addition,it is beneficial for policy management that using Hash table to store the mappings between textuary attributes and numerical attributes.Simulation experimental results show that the policy engine using the policy optimization method proposed in this paper is much faster than Sun XACML.
Algebraic Side-channel Attacks Method of ITUbee
LI Lang and DU Guo-quan
Computer Science. 2016, 43 (2): 169-174.  doi:10.11896/j.issn.1002-137X.2016.02.037
Abstract PDF(434KB) ( 457 )   
References | Related Articles | Metrics
ITUbee was proposed in the second lightweight cryptography for security and privacy 2013.It has great significance to do security analysis about ITUbee.The algebraic side-channel attacks methods of ITUbee were researched.First,we constructed the equivalent-algebraic equations of ITUbee S-box.But,it is difficult to work out the structured equations set.The leakage of cryptographic power consumption of ITUbee algorithm was collected.The Hamming weight of the encryption middle status byte was inferred.Then,the simultaneous Boolean equations set with the cipher algorithm was conversed.At last,we used the cryptominisat to solve the key.Experiment results show that it only needs less samples to gain the successful attack.The initial keys can be derived via analyzing the part HW (Hamming weight) leakages of the first round in the scene of the known-plaintext and the unknown ciphertext.
Ciphertext-policy Attribute-based Encryption with Anonymous Access Structure
WANG Hai-ping and ZHAO Jing-jing
Computer Science. 2016, 43 (2): 175-178.  doi:10.11896/j.issn.1002-137X.2016.02.038
Abstract PDF(415KB) ( 624 )   
References | Related Articles | Metrics
In ciphertext-policy attribute-based encryption(CP-ABE) scheme,a user’s secret key is associated with a set of attributes,and the ciphertext is associated with an access policy.The user can decrypt the ciphertext if and only if the attribute set embedded in his secret key satisfies the access policy specified in the ciphertext.In the present schemes,the access policy is sent to the decryptor along with the ciphertext,which means that the privacy of the encryptor is revealed.In order to solve such problem,we proposed a CP-ABE scheme with anonymous access policy,which is able to preserve the privacy of the encryptor.Our new scheme is proved to be selectively secure against chosen-plaintext attack under DBDH assumption in the standard model.
Correctness Analysis and Improvement of Group Signature in Composite Order Bilinear Groups
YU Jia-fu, ZHONG Hong and WANG Yi-min
Computer Science. 2016, 43 (2): 179-182.  doi:10.11896/j.issn.1002-137X.2016.02.039
Abstract PDF(367KB) ( 379 )   
References | Related Articles | Metrics
Zhou Fu-cai et al proposed an efficient group signature scheme based on BMW model by utilizing the composi-te order bilinear groups theory and non-interactive zero knowledge proof system.However,this study demonstrates that there are some deficiencies in Zhou’s scheme that signature verifier cannot verify signer’s ID correctly and cannot finish the signature verification.Then,the authors provided an improved scheme and proved its security strictly.The proposed scheme corrects the errors by adding the commitment to signer’s ID and corresponding non-interactive zero knowledge proof.At last,this paper compared the security and efficiency respectively with the similar group signatures.And the result of analysis shows that the improved scheme resolves the problem of Zhou’s scheme in the premise of assuring the security and efficiency.
Checking Topological Integrity of Geographic Data Based on Fragile Watermarking
SUI Li-li and WANG Chuan-jian
Computer Science. 2016, 43 (2): 183-187.  doi:10.11896/j.issn.1002-137X.2016.02.040
Abstract PDF(395KB) ( 342 )   
References | Related Articles | Metrics
Topological relation is the basis of spatial relationships,which is widely used in spatial query and spatial reasoning.In this paper,a fragile watermarking method was proposed to verify the topological integrity of geographic data.The watermark is generated from the disjoint distance between the objects.Then the objects are zoomed in or out according to the modified disjoint distance rate based on the watermark to embed the watermark.In watermarking detection,the topological integrity of geographic data is determined by the result of matching the regenerated watermark with the extracted watermark.Experimental results demonstrate that it is effective to verify topological integrity of geographic data by the method.
Performance Analysis of PHY-CRAM Physical Layer Challenge Response Authentication Mechanism
ZHANG Dan, WU Xiao-fu, YAN Jun and ZHU Wei-ping
Computer Science. 2016, 43 (2): 188-191.  doi:10.11896/j.issn.1002-137X.2016.02.041
Abstract PDF(384KB) ( 439 )   
References | Related Articles | Metrics
The physical layer authentication recognizes the identities of communication users by exploiting the characteristics of physical layer resources,which can effectively prevent illegal users from accessing the wireless network in order to enhance the information security of wireless network.PHY-CRAM proposed recently is a typical physical layer challenge response authentication mechanism,and its performance mainly depends on the simulation.So this paper attempted to theoretically analyze the authentication performance and derived the analytical expressions of successful authentication rate and false acceptance rate.The derived results show that the correlation coefficient of PHY-CRAM authentication standards obeys Rice distribution,and successful authentication rate and false acceptance rate can be calculated by Marcum Q function.The computer simulation results show that the probability density curve of correlation coefficient is coincident with that of Rice distribution,and the longer the secret key is,the better the fitness is.The theory value of receiver operating characteristic curve is identical with the statistical value.
UML Model to Simulink Model Transformation Method in Design of Embedded Software
GUO Peng, LI Ya-hui, SUN Lei and CAI Xiao-le
Computer Science. 2016, 43 (2): 192-198.  doi:10.11896/j.issn.1002-137X.2016.02.042
Abstract PDF(838KB) ( 600 )   
References | Related Articles | Metrics
Model driven development and its key technique model transformation are research hotspot of software engineering in recent years.At the early stage of embedded software development,design model not only requires static analysis,but also needs dynamic simulation,verifying correctness of system design.How to transform design model to simulation model is a serious problem to industrial department.This paper surveyed model transformation research status,analysed related model transformation techniques of model drive development,proposed a model transformation method from UML to Simulink,built UML meta-model and Simulink meta-model,designed a set of mapping rule bet-ween UML meta-model and Simulink meta-model.Finally,this paper validated technique and method correctness using automatic flight control system as antitype.The method makes two isomerism models homogeneous,improving the efficiency of embedded software development,enriching MDD technique,and providing technique support for embedded software development,such as automobile control system,express control system,and avionics system.
Dynamic Mutation Execution Strategy for Mutation-based Fault Localization
GONG Pei, GENG Chu-yao, GUO Jun-xia and ZHAO Rui-lian
Computer Science. 2016, 43 (2): 199-203.  doi:10.11896/j.issn.1002-137X.2016.02.043
Abstract PDF(513KB) ( 384 )   
References | Related Articles | Metrics
During software debugging process,how to locate faults in programs quickly and accurately is an issue of common concern to developers.Mutation-based fault localization is an approach by estimating the possibilities of statements that incur error to locate faults on the basis of the similarity between the programs under test and corresponding mutants.This approach shows a high precision on fault localization but requires a large execution cost,since it needs to execute the test suite on a lot of mutants.For reducing unnecessary execution cost,this paper presented a dynamic mutation execution strategy,which dynamically adjusts the execution orders of both mutants and test cases according to previous execution information.Empirical studies were conducted on 127 faulty versions from 6 program packages.The results indicate that this strategy can reduce 23% up to 78% mutation execution cost under the case of keeping fault location precision.
Fuzzy Comprehensive Evaluation Method Based on Measure of Medium Truth Degree
XU Wen-hua, CHEN Hai-yan, ZHANG Yu-ping and WANG Jian-dong
Computer Science. 2016, 43 (2): 204-209.  doi:10.11896/j.issn.1002-137X.2016.02.044
Abstract PDF(498KB) ( 335 )   
References | Related Articles | Metrics
Aiming at the problems that most of the existing comprehensive evaluation methods give rigid evaluation results,and they do not apply to the evaluation of objects with fuzzy factors,a comprehensive evaluation method based on measure of medium truth degree was proposed.The processes of the measure of medium truth degree and the methods of setting the parameters were designed respectively when different types of indexes,such as beneficial index,cost index and interval index,were given.Afterwards,the intermediary fuzzy evaluation matrix was constructed based on the mea-surement results of medium truth degree.Intermediary mathematical system was created for processing fuzzy information.Compared to the traditional fuzzy evaluation method,the comprehensive evaluation method based on the measure of medium truth degree is more objective so that it is able to deal with fuzzy phenomenon more effectively.The comparative experimental results indicate that compared to other evaluation methods,the proposed method is valid and has certain advantages.
Structure Optimization for Automatic Vectorization
YU Hai-ning, HAN Lin and LI Peng-yuan
Computer Science. 2016, 43 (2): 210-215.  doi:10.11896/j.issn.1002-137X.2016.02.045
Abstract PDF(1029KB) ( 588 )   
References | Related Articles | Metrics
Sturcture is used more extensively to promote the performance of application program such as scientific computing.The noncontinuity and the nonalignment of its non-array memory address have a dramatic influence on the efficiency of program’s vectorization.To reduce the access to these addresses during the SIMD’s vectorization,this paper applied a structure peeling model based on the structure which combines field access affinity with type to eliminate the “clearance” of memory between field storage,and proposed an address conversion method of structure array one by one mapping to the two dimensional array to meet the request of the continuity and the alignment of its non-array memory address,further reducing the failure rate of Cache,so as to improve application performance.By using the test suites of gcc-vec,spec2000 and spec2006,the experimental results on the compiler of automatic vector show that using the me-thod,the performance of optimized programs can be improved by more than eight percent.
Method of Code Evolution Recommendation for Database Schema Change
ZHANG Wu-neng, LI Hong-wei, SHEN Li-wei and ZHAO Wen-yun
Computer Science. 2016, 43 (2): 216-223.  doi:10.11896/j.issn.1002-137X.2016.02.046
Abstract PDF(1175KB) ( 366 )   
References | Related Articles | Metrics
Many software store information relies on database,while the database schema change will lead to some source code SQL statements associated with the database cannot be performed properly.Thus finding a way to locate the SQL statement code which needs modifying and to recommend possible changes to these programs is necessary.We proposed a code evolution recommendation method for database schema changes.Firstly,the method detects software system database schema change,and uses the program slicing technique to get source code fragments related to the ope-ration of the database.Then it determines the program slicing impacted by the database schema change,and uses algorithm for generating flowchart from source program to get program flowcharts.SQL statements may have some distinct execution paths according to program flow branches condition.Finally,it uses figure mapping method for each SQL statement to recommend executable SQL statements under the new database schema.In order to verify the feasibility of the method,the article implemented a plug-in tool used to detect the database schema changes,and to recommend executable SQL statement under new database schema.
Traffic Speed Forecasting Method Based on Nonparametric Regression
SHI Dian-xi, DING Tao-jie, DING Bo and LIU Hui
Computer Science. 2016, 43 (2): 224-229.  doi:10.11896/j.issn.1002-137X.2016.02.047
Abstract PDF(511KB) ( 349 )   
References | Related Articles | Metrics
Non-parametric regression model is a traffic forecasting model proposed in recent years.Based on the characteristics of the model,in order to improve the forecasting precision on the issue of neighboring states selection,the original neighbor matching was optimized by the classification of the trend of speed and varying K neighbors precise search strategy based on intensity,and then a short-term traffic speed forecasting model was proposed.Floating car data in Beijing was used in the experiments.Results show that the optimized model is better than normal non-parametric regression model and BP neural network model,and can provide practical speed for short-term traffic prediction.
Incremental Learning Algorithm Based on Twin Support Vector Regression
HAO Yun-he and ZHANG Hao-feng
Computer Science. 2016, 43 (2): 230-234.  doi:10.11896/j.issn.1002-137X.2016.02.048
Abstract PDF(410KB) ( 434 )   
References | Related Articles | Metrics
This paper proposed an incremental learning algorithm based on twin support vector regression.When a new sample is added to the training set,our algorithm makes full use of old computing information instead of training all the new training set,so it greatly simplifies the calculation of inverse matrix and improves the execution efficiency.Experimental results on artificial datasets,time series and UCI datasets show that our algorithm has remarkable improvement of generalization performance with short training time.
Target Grouping Algorithm Based on Multiple Combat Formations
YUAN De-ping, ZHENG Juan-yi, SHI Hao-shan and LIU Ning
Computer Science. 2016, 43 (2): 235-238.  doi:10.11896/j.issn.1002-137X.2016.02.049
Abstract PDF(394KB) ( 566 )   
References | Related Articles | Metrics
A clustering algorithm was proposed for grouping enemy targets in the situation of the enemy multiple targets attacking our target groups.Firstly,the spatial clustering is realized by the constrained chameleon algorithm based on the geometric elements of the enemy’s targets.And then,the advantage function of the attacking elements for the ene-my’s space group is calculated by the geometric elements of enemy space groups,and the attack factor matrix of the space groups between two sides is formed.Finally,the enemy’s relationship groups are divided by a series derivation including computing the subjective weight and objective weight of the attack elements,deducing the synthetic weight and the attack matrix of the space groups between two sides.The effectiveness of the proposed algorithm was verified by the simulation of the given scene.
Liheci Word Sense Disambiguation Based on SVM
ZHANG Zhen-jing, LI Xin-fu, TIAN Xue-dong and WANG Kai
Computer Science. 2016, 43 (2): 239-244.  doi:10.11896/j.issn.1002-137X.2016.02.050
Abstract PDF(506KB) ( 558 )   
References | Related Articles | Metrics
The task of Liheci word sense disambiguation is to make computers choose the correct sense of a Liheci ambiguous word in a given context.For the problem that a Liheci ambiguous word in machine translation is not accurate and in the information retrieval is unable to match the useful information,a word sense disambiguation method was applied to the Liheci ambiguous words and a classifier model was established using SVM.In order to improve the accuracy of the Liheci word sense disambiguation,it extracts not only local word,local part of speech,local word and part of speech,but also the middle insert part of the separated form as disambiguation features according to the characteristics of Liheci.When the text characteristics was converted to feature vector,we could fixed feature weights of some type in turn and changed the feature weights of the other two types to verity the disambiguation effect of the three kinds of feature,respectively.The results show that the effect of local word feature,local word and part of speech features on disambiguation is higher than local part of speech,and using different types of feature weight compared with the same,disambiguation accuracy increases by 1.03%~5.69%.
DTW Clustering-based Similarity Mining Method for Hydrological Time Series
YANG Yan-lin, YE Feng, LV Xin, YU Lin and LIU Xuan
Computer Science. 2016, 43 (2): 245-249.  doi:10.11896/j.issn.1002-137X.2016.02.051
Abstract PDF(384KB) ( 608 )   
References | Related Articles | Metrics
Similarity mining of hydrological time series is an importance aspect of hydrological time series mining.It will be of great importance in flood forecasting and flood control scheduling.According to the characteristics of hydrological data,this paper proposed a DTW clustering-based similarity mining method over hydrological time series.Firstly,on the premise of wavelet denoising,feature point segmentation and semantic classification,hierarchical cluster analysis is used to the classified sub-sequences based on DTW distance and the sub-sequences are symbolized.Then,candidate sets of time series are filtered according to the edit distance between symbol sequences.Finally,the similar hydrological time series are got precisely from the candidate sets by DTW exact matching.Experiments on the water level of Chuhe Liuhe station show that the proposed method can narrow the candidate sets effectively and improve the efficiency of searching for semantic similarity of hydrological time series.
Outer Inverse P-sets and Double Information Camouflage-Recovery for Digital Image and its Application
ZHANG Jing-xiao, XU Feng-sheng, LIU Xin-hua and ZHANG Li-hua
Computer Science. 2016, 43 (2): 250-253.  doi:10.11896/j.issn.1002-137X.2016.02.052
Abstract PDF(380KB) ( 371 )   
References | Related Articles | Metrics
This paper first introduced operator into study of outer inverse P-sets and broaden the application of inverse P-sets.Then the theory of outer inverse P-sets was applied in information camouflage of digital image,and the concepts such asf- information camouflage,g- information camouflage,double information camouflage and information camouflage metric were established.The information camouflage metric property theorem,f- information camouflage-recovery theorem,g- information camouflage-recovery theorem of digital image and double information camouflage-recovery theorem of digital image were proposed.Finally,the application example was given.
Semantic-based Feature Extraction Method for Document
JIANG Fang, LI Guo-he and YUE Xiang
Computer Science. 2016, 43 (2): 254-258.  doi:10.11896/j.issn.1002-137X.2016.02.053
Abstract PDF(411KB) ( 389 )   
References | Related Articles | Metrics
Feature extraction of Chinese documents is an important part in the document processing,and imposes great influence on the document classification.Pre-existing document feature extraction methods have many shortcomings,such as creating a feature vector of high dimensions,depending on training sets,ignoring low-frequency keywords,and so on.In this paper,the semantic distance between words was calculated based on the synonyms dictionary,and then theme related words of each classification were selected by the density clustering method,and finally the feature words were selected from the theme related words using the information gain algorithm.In order to validate the proposed method,one validation experiment and one comparison experiment were designed and the evaluation indexes including the macro-F value and the micro-F value were calculated.Experiment results show that the proposed document feature extraction method has better performance than other traditional algorithms.
Research of Heart Rate Variable Analysis Based on Sliding Window Hurst
LV Tai-zhi
Computer Science. 2016, 43 (2): 259-262.  doi:10.11896/j.issn.1002-137X.2016.02.054
Abstract PDF(565KB) ( 388 )   
References | Related Articles | Metrics
Heart rate variability (HRV) is the physiological phenomenon of variation in the time interval between heartbeats,and HRV analysis can be used as a diagnosis method for assessing the physiological and psychological states.Up to now,most HRV analyse have been done offline and only have being applied in clinical application and research.The paper proposed a real-time HRV signal sampling and analysis system based on Android platform.This system uses the IOIO board,Wifi or bluetooth to make a connection between Android devices and mobile or wearable health sensors.This paper used slide window based Hurst exponent series to analyze the sampling data.It uses two indices,the cumulative mean of Hurst series (CMHurst) and the cumulative standard deviation of Hurst series (CStdHurst),to estimate the heart health status.The indices were calculated from the windowed estimated Hurst series.To verify the validity of this method,the indices were tested by some databases from PhysioBank.The experiment results show this method can distinguish the groups who have normal rhythm or abnormal rhythm.
Accelerating Structure Learning of Bayesian Network
SEIN Minn and FU Shun-kai
Computer Science. 2016, 43 (2): 263-268.  doi:10.11896/j.issn.1002-137X.2016.02.055
Abstract PDF(570KB) ( 562 )   
References | Related Articles | Metrics
Structure learning is the basis for the application of Bayesian networks (BN).A novel algorithm called APC was proposed to recovery the whole structure via sequential induction of local structures.APC inherits the most feature of PC algorithm,i.e.effectively avoiding high-dimensional conditional independence (CI) tests.Besides,it constructs and sorts candidate sets which possibly d-separate any pair of nodes,X and Y,based on information implied in early conducted CI tests and known features of BN topology.Then,CI tests involving highly ranked candidate set are performed with priority.This strategy is expected to avoid fruitless CI tests,and up to 50% saving is observed on APC over PC in our experimental study.
Rough Set Model Based on Logical And Operator and Logical Disjunct Operator of Variable Precision and Grade in Ordered Information System
YU Jian-hang and XU Wei-hua
Computer Science. 2016, 43 (2): 269-272.  doi:10.11896/j.issn.1002-137X.2016.02.056
Abstract PDF(268KB) ( 288 )   
References | Related Articles | Metrics
In this paper,the rough set model based on logical and operator and logical disjunct operator of variable precision and graded was proposed in ordered information system.The new model is an expansion of the classical rough set,variable precision rough set and graded rough set model.The accurate description of this model was discussed and some important properties of this model were investigated carefully.Moreover,a specific case study about the student achievement was analyzed.This study provides a new way for the knowledge discovery in ordered information system.
New Note on Classical Measure of Knowledge Dependency and Attribute Significance
CHEN Fei, JIANG Lin and LI Jin-hai
Computer Science. 2016, 43 (2): 273-276.  doi:10.11896/j.issn.1002-137X.2016.02.057
Abstract PDF(382KB) ( 403 )   
References | Related Articles | Metrics
Measure of knowledge dependency and attribute significance is an important issue in rough set theory.It has been widely applied to knowledge reduction,rule extraction,etc.Classical measure of knowledge dependency and attri-bute significance has a littie bit of limitation in dealing with data,and sometimes it cannot obtain precise and reasonable results,which leads to a series of deviations in the subsequent applications.To this end,through deep analysis of the exi-sting classical knowledge dependency,combining it with the majority inclusion relation,and adding credibility parameters,a new method of measuring knowledge dependency and attribute significance was proposed.Finally,the new mea-sure method was applied to decision information system.And the analysis results show that the new method is effective.
Building User Personalization Model Based on Short Term Memory and Forgetting Factor
CHEN Hai-yan, XU Zheng and ZHANG Hui
Computer Science. 2016, 43 (2): 277-282.  doi:10.11896/j.issn.1002-137X.2016.02.058
Abstract PDF(774KB) ( 341 )   
References | Related Articles | Metrics
One criticism of search engines is that when queries are submitted,the same results are returned to different users.To address the accuratcy problem,personalized search was proposed,since it could provide different search results based upon the preferences of users.However,the existing methods concentrate more on the long-term and independent user profile,thus reducing the effectiveness of personalized search.We introduced an approach that captures the user context to provide accurate preferences of users for effectively personalized search.First,the short-term query context is generated to identify related concepts of the query.Second,the user context is generated based on the click through data of users.Finally,a forgetting factor is introduced to merge the independent user context in a user session,which maintains the evolution of user preferences.The experimental results fully confirm that our approach can successfully represent user context according to individual user information needs.
PID-type Iterative Learning Control for a Class of Nonlinear Systems with Arbitrary Initial Value
HAO Xiao-hong, LI Zhuo-yue and WANG Hua
Computer Science. 2016, 43 (2): 283-286.  doi:10.11896/j.issn.1002-137X.2016.02.059
Abstract PDF(270KB) ( 516 )   
References | Related Articles | Metrics
The paper gave an PID-type iterative learning control algorithm with initial state study aiming at a class of nonlinear system running repeatedly in limited interval.Operator theory proves the validity of the algorithm.The algorithm not only solves initial state problem of the iterative learning control,but also relaxes the convergence conditions.Results of comparing the simulation analysis of PID-type iterative learning control algorithm with the simulation analysis of PD-type iterative learning control algorithm show that the tracking accuracy of nonlinear system in the initial state conditions is significantly improved after PID-type iterative learning,and the output error curve tends to zero more quickly.The results illustrate the effectiveness of the algorithm.
Multi-instance Multi-label Learning Algorithm by Treating Instances as Non-independent Identically Distributed Samples
CHEN Tong-tong, DING Xin-miao, LIU Chan-juan, ZOU Hai-lin, ZHOU Shu-sen and LIU Ying
Computer Science. 2016, 43 (2): 287-292.  doi:10.11896/j.issn.1002-137X.2016.02.060
Abstract PDF(485KB) ( 722 )   
References | Related Articles | Metrics
Multi-instance multi-label learning (MIML) is a new machine learning framework.In this framework,an object is represented as a bag which is decomposed of multiple instances and labeled with multiple labels.Previous studies on MIML typically treated instances in the bags are independently identically distributed.However,it is difficult to be guaranteed in real tasks.Considering correlation features of instance in a bag,a multi-instance multi-label learning algorithm by treating instances as non-independent identically distributed samples was proposed.Firstly,instance correlations are considered by establishing an affinity matrix.By this means each bag is represented with an affinity matrix.Then,kernel functions based on the affinity matrix in different scales are established.Finally,considering predictions of different kinds of labels are corresponding to different kernels,multiple kernel learning is introduced to construct and train the MKSVMs.Experimental results on image data set and text data set show that the proposed algorithm greatly improves the accuracy of the image multi-label classification compared with other methods.
Co-location Patterns Mining with Time Constraint
ZENG Xin and YANG Jian
Computer Science. 2016, 43 (2): 293-296.  doi:10.11896/j.issn.1002-137X.2016.02.061
Abstract PDF(388KB) ( 399 )   
References | Related Articles | Metrics
Most of the research achievements of spatial data mining are based on the ideal spatial data and the idea of examples equality,ignoring the time constraint condition existing in the actual scene.This paper considered the existent time interval of the instance as constraint condition,redefined spatial neighborhood relation R,proposed spatial frequent pattern mining algorithm TI with time constraint,and by using time overlap as pruning condition,proposed pruning algorithm TI-C.Through empirical data analysis,under the same data set,the efficiency of TI-C algorithm is better than that of TI,the frequent pattern number of TI-C algorithm is less than that of join-based algorithm,and the frequent pattern of TI-C algorithm can accurately and truly reflect the object’s co-location relation of the actual scene.
Method about Reconstruction of SEM Image Based on Gradient
MAO Xiang-di and SHI Zheng
Computer Science. 2016, 43 (2): 297-301.  doi:10.11896/j.issn.1002-137X.2016.02.062
Abstract PDF(1473KB) ( 329 )   
References | Related Articles | Metrics
Aiming at the problem about binaryzation of gray level image which is produced by scanning electron microscope(SEM) in the process of integrated circuit manufacturing,this paper put forward a method about reconstruction based on the gradient of SEM image and statistics.Otsu is used to analyze the noise of SEM image,and by means of smoothing,the gradient information of image is got with Kirsch operator.Then according to the feature that the gra-dient of outer edge is larger than inner edge,the algorithm uses statistic information to fill the image.Experimental results show that the algorithm has high stability and automation in high resolution images,and in low resolution images,it avoids the effect of edge extracting failing and produces correct binary image completely.
Video Segmentation Algorithm Based on Join Weight of Superpixels
SUN Tao and CHEN Kang-rui
Computer Science. 2016, 43 (2): 302-306.  doi:10.11896/j.issn.1002-137X.2016.02.063
Abstract PDF(660KB) ( 406 )   
References | Related Articles | Metrics
Video segmentation is a hot issue in the field of image processing.Based on traditional segmentation algorithms,a new unsupervised video segmentation algorithm was proposed.This algorithm represents the moving foreground with superpixel algorithm,defines the join weight of superpixels as the possibility from the same object,and calculates the join weight with static features from current frame associated with the relevance feature between frames.In order to optimize the search of relevance match between superpixels from different frames,the algorithm introduces superpixel color feature constraint and movement constraint.The experiment contains two aspects,and the algorithm ensures higher recall rate and stable precision rate in the simple scenario and completes single person segmentation from the crowd in the complex scenes.Large numbers of experiments show that the proposed algorithm can realize video image segmentation,and effectively solve the problem of over-segmentation.
Enhanced Block Compressed Sensing of Images Based on Total Variation Using Texture Information
WANG Yue, ZHOU Cheng, XIONG Cheng-yi and SHU Zhen-yu
Computer Science. 2016, 43 (2): 307-310.  doi:10.11896/j.issn.1002-137X.2016.02.064
Abstract PDF(700KB) ( 350 )   
References | Related Articles | Metrics
Block compressed sensing of images solves the problems of high computational complexity and large storage space required by blocking an image and downsizing measurement matrix.But such a practice will result in blocking artifacts,which needs to be filtered.Existing algorithms do not consider how to recover textural features of images,which will result in quality degradation of image reconstruction.In order to solve this problem,this paper proposed an algorithm which uses an adaptive sampling model based on gray entropy at first,and then analyzed the reason why blocking artifacts generate and are reduced by adaptive sampling.At last,in the proposed algorithm TV filter is joined with SPL process,and a DDWT/TV filter model based on texture information is built to replace the former filtering process in reconstruction.The model can preserve more details of images after decreasing block artifacts by using adaptive sampling.Experimental results show that the proposed algorithm can remarkably improve the subjective and objective quality of the reconstructed image and can effectively hold more texture information of images compared to some existing methods.
Chemotaxis Operator Embedded Particle Swarm Optimization Algorithm and its Application to Multilevel Thresholding
ZHANG Xin-ming, TU Qiang, YIN Xin-xin and FENG Meng-qing
Computer Science. 2016, 43 (2): 311-315.  doi:10.11896/j.issn.1002-137X.2016.02.065
Abstract PDF(705KB) ( 325 )   
References | Related Articles | Metrics
The standard particle swarm optimization (PSO) algorithm is easy to trap into local optimum when selecting the optimal thresholds in multilevel thresholding,so a novel PSO algorithm by embedding the chemotaxis operator was presented.The standard PSO algorithm often possesses the strong global search ability but poor local search ability,while the feature of bacterial foraging optimization (BFO) is just reverse.The BFO’s chemotaxis operator with good local search ability is embedded into the PSO,and the chemotaxis operator embedded PSO (COPSO) algorithm is got.On the basis of complementary advantages,the COPSO has both good global search ability and local search ability.The optimal threshold vectors can be obtained by applying the COPSO algorithm to multilevel image thresholding based on maximum entropy.The experimental results demonstrate that the COPSO algorithm can get better optimization effect and shorter optimization time compared with standard PSO,BFO and GA.
Study of Handwriting Recognition of Improved Algorithm Based on PCNN
Computer Science. 2016, 43 (2): 316-318.  doi:10.11896/j.issn.1002-137X.2016.02.066
Abstract PDF(521KB) ( 366 )   
References | Related Articles | Metrics
Pulse coupled neural network(PCNN)is widely used in image processing,pattern recognition and other fields.This paper presented an improved algorithm for foveation points detection based on PCNN.First,neuron excitation function is improved,and wavelet shrinkage method is applied in the image to reduce the noise,preserving the hie-rarchy of image.Then the handwriting is identified through the foveation points detection.The experimental results indicate that the method can effectively improve the recognition rate of handwritten letters.Especially in the noise environment,the recognition rate can be greatly improved.