Started in January,1974(Monthly)
Supervised and Sponsored by Chongqing Southwest Information Co., Ltd.
ISSN 1002-137X
CN 50-1075/TP
CODEN JKIEBK
Editors
Current Issue
Volume 43 Issue 9, 01 December 2018
  
Review of 3D Stereo Vision Measure and Reconstruction Based on Mirror Image
GUO Wei-qing, TANG Yi-ping, LU Shao-hui and CHEN Qi
Computer Science. 2016, 43 (9): 1-10.  doi:10.11896/j.issn.1002-137X.2016.09.001
Abstract PDF(5839KB) ( 2232 )   
References | Related Articles | Metrics
The 3D information of the object or scene obtained by using the mirror image is getting more and more attention from many researchers.Mirror image occurs in the interaction between light and plane/curved mirror.The reflective properties of plane mirror image can improve our visual effects.Ray unfolding procedure can be used in different mirror image systems.Using ray unfolding instead of mirror interaction is applied in 3D scene,to get a virtual 3D space.Plane mirror image enables complex ray interactions to be visualized in a virtual way,and the changes of coordinate system are easy to be traced.Curved mirror image usually does not have the properties of perspective projection,and it changes the display space according to the curvature of the surface.Curved mirror often leads to refraction and reflection,hence it is necessary to design the corresponding algorithm of geometric restoration for different 3D stereo vision measure and reconstruction.Form the perspective of computer graphics and computer vision,this paper analyzed the basic principle of mirror image,reviewed the typical 3D measurement and reconstruction method based on mirror image technology in recent years and the latest research progress.
Review of Remote Sensing Image Classification Based on Support Vector Machine
WANG Zhen-wu, SUN Jai-jun,  YU Zhong-yi and BU Yi-ya
Computer Science. 2016, 43 (9): 11-17.  doi:10.11896/j.issn.1002-137X.2016.09.002
Abstract PDF(740KB) ( 683 )   
References | Related Articles | Metrics
Remote sensing technology is an important technology of studying the earth mineral resources and energy.Remote sensing image classification plays a key role in the application of remote sensing technology.Support vector machine(SVM) is a machine learning method based on VC dimension(Vapnik-Chervonenkis Dimension) theory and structural risk minimization principle,which has been widely used in the actual remote sensing image classification.Domestic and foreign scholars have done a lot of research about it and these studies were systematically summarized in this paper.The remote sensing image classification methods based on the support vector machine is reviewed hierarchically,that is,the principle and characteristics of each method were analyzed and compared laterally and vertically.The research status of the remote sensing image classification based on the support vector machine was summarized systematically and completely in this paper.Finally,the future development direction of support vector machine algorithm applied in the remote sensing image classification was pointed out.
Complexity of Flow Shop Scheduling Problems with No-load Return Transportation
LAN Yan, ZHANG Ming-hui, WU Zong-tao and HAN Xin
Computer Science. 2016, 43 (9): 18-22.  doi:10.11896/j.issn.1002-137X.2016.09.003
Abstract PDF(1632KB) ( 461 )   
References | Related Articles | Metrics
Flow shop is one of the most classic scheduling problems.The model of a flow shop problem with two machines and a transporter which can transport a job from machine 1(M1) to machine 2(M2) every time is widely applied in most manufacturing systems.In this paper,the no-load return transportation time from machine M2 to M1 was consi-dered,and we assumed that it equals with the time from M1 to M2.Then we showed that the NP-complete problem 3-PARTITION is reducible to an instance of the problem we researched,so that we can demonstrate that this problem is a NP-hard problem in the strong sense.
Research on Complexity and Approximation Algorithm for Counting 3-Set Packings of Size k
LIU Yun-long
Computer Science. 2016, 43 (9): 23-26.  doi:10.11896/j.issn.1002-137X.2016.09.004
Abstract PDF(315KB) ( 1110 )   
References | Related Articles | Metrics
Counting 3-Set Packings of size k is to count distinct packings of size k in a given instance of 3-Set Packing.We first showed that the complexity of this problem is #W[1]-hard,which indicates that there exists no efficient fixed-parameter tractable algorithm for it(unless #W[1]=FPT).Subsequently,by extending the algorithm for counting 3-D Matchings of size k,we obtained a generalized approximation algorithm for counting 3-Set Packings of size k.This algorithm is heavily based on the Monte-Carlo self-adjusting coverage algorithm and the recent improved color-coding techniques.
Join Algorithm in Skewed Datasets Based on MapReduce
LIANG Jun-jie and HE Li-min
Computer Science. 2016, 43 (9): 27-31.  doi:10.11896/j.issn.1002-137X.2016.09.005
Abstract PDF(1693KB) ( 687 )   
References | Related Articles | Metrics
Join operation is the most common operation in data analysis applications with large-scale datasets,and Map-Reduce can not support join operation perfectly in handling data skew problem.MapReduce frequecncy classified join algorithm was proposed,and datasets were classified into three categories according to the appeared data frequency.Data redistribution applying partitioning algorithm and broadcast algorithms eliminate the impact of skewed data.And data redistribution is realized by using hash algorithm for the non-skew data.Join operation can be completed in a single node,avoiding the cost of communications across the nodes under the MapReduce for the redistributed data,and balancing the workload of each node effectively,thereby improves the efficiency of join operations in skewed data.The effectiveness and practicality of the algorithms are proved by the comparison with traditional algorithms.
Memory Computing:Concept,Characteristics and Research Advances
GUO Bin, CHEN Hui-hui, LI Wen-peng, YU Zhi-wen, JIANG Jia-jun and WANG Wen-hui
Computer Science. 2016, 43 (9): 32-38.  doi:10.11896/j.issn.1002-137X.2016.09.006
Abstract PDF(603KB) ( 475 )   
References | Related Articles | Metrics
With the development of information and communication technology,especially the mobile Internet and the Internet of Things,the data about personal works and life grow exponentially.Since much valuable individual information is invisibly contained in these massive dataset.How to extract the valuable information from these great amount of data has become a new challenge.This paper introduced an newly emerging research topic in the pervasive computing field,memory computing.By utilizing prevalent sensor-embedded smartphones and wearable devices,memory computing collects heterogeneous online/offline data of user’s daily activities,extracts and mines inner memory data and associated contexts,manages useful data in the right manner,and renders context-aware individual memory information for helping users to recall their memories and promote their communication and cooperation.This paper presented the concept,features and the system model of memory computing,and discussed its key technologies and research challenges.We also reviewed the applications of memory computing,such as life logging,reminiscence,memory reminding and community memory sharing,and prospected the future development.
FP-CNNH:A Fast Image Hashing Algorithm Based on Deep Convolutional Neural Network
LIU Ye, PAN Yan, XIA Rong-kai, LIU Di and YIN Jian
Computer Science. 2016, 43 (9): 39-46.  doi:10.11896/j.issn.1002-137X.2016.09.007
Abstract PDF(1417KB) ( 1458 )   
References | Related Articles | Metrics
In the big data era,application research on image retrieval technology for large-scale data is a hot field.Recent years,image hash algorithm has attracted much attention in large-scale image retrieval system in order to improve the retrieval efficiency and reduce the storage space.However,there are some issues with existing supervised hash code learning algorithms.Most of the supervised hash algorithms need to use image feature extractor for obtaining hand-crafted image features,which influence the effect of hash code training with image features,at the same time these methods cannot deal well with semantic similarity for image data set.With the development of deep learning research on the large-scale data,some recent related work try to deploy deep neural network to learn hash function and make the hash code training effect increased.But such kind of methods require carefully designed complex neural network structure thus increase the difficulty of the hash function design and cost more time on neural network training with large data set.These problems limit the range of the hash algorithm application with the deep learning architecture for large data sets.To solve the problems mentioned above,this paper proposed a fast image hashing algorithm based on deep convolution neural network.The proposed algorithm consists of an optimization approach for constructing the hash code of the training data set and a pre-trained large deep neural network for learning to improve the effect of hash algorithm,shor-tening the training time of complex neural network.According to the analysis of experimental results on different image data sets,both the effectiveness of hash function and the efficiency of training time of the proposed algorithm have better performance compared with the existing algorithms.
Online Learning Based on Stochastic Spectral Gradient
XUE Wei, ZHANG Wen-sheng and REN Jun-hong
Computer Science. 2016, 43 (9): 47-51.  doi:10.11896/j.issn.1002-137X.2016.09.008
Abstract PDF(344KB) ( 653 )   
References | Related Articles | Metrics
We considered a type of learning problems whose objective functions can be formulated as an average of a large number of component functions,and assumed that each component function is smooth.Among many machine learning methods,online learning provides a favorable tool for big data learning,not only due to its simple operation and fast convergence rate,but also for its ability to automatically update model.To solve such problems,we developed a stochastic spectral gradient descent (S2GD) method,which employs the Rayleigh quotient to collect second-order information to construct Hessian inverse approximations.S2GD can be viewed as an approach that extends the spectral gradient method working in deterministic setting to stochastic setting.At each iteration,the generated search direction guarantees descent property.The existing conclusion indicates that the S2GD method is of convergence.Preliminary experimental results on LIBSVM data sets are reported to demonstrate the feasibility and effectiveness of our approach.
Construction Method of Academic Circle Based on Label Similarity Computation
FU Cheng-zhou, TANG Yong, HE Chao-bo, WANG Jin-ling and YUAN Cheng-zhe
Computer Science. 2016, 43 (9): 52-56.  doi:10.11896/j.issn.1002-137X.2016.09.009
Abstract PDF(495KB) ( 460 )   
References | Related Articles | Metrics
Constructing academic circles for users in the scholar-oriented social network system has important application values for promoting exchanges among scholars.Similarity computation is done based on the common properties of scholars,constituting academic circles with similar academic field and research subject,allowing scholars to collaborate more closely and frequently.This paper proposed a method which uses scholars of the academic information to extract the personal characteristics of academic labels,and measures the weight of different class labels.Then through the similarity computation and clustering algorithm,the academic circle can be constructed.By crawling the public information on the academic social network platform named SCHOLAT to perform experiments,the reliability and usefulness of the proposed method are verified.
Research on Studying Method of Network Anomalous Behaviors Classification Based on Topic Model
MA Zheng-ran, ZHANG Bo-feng and WANG Yong-jun
Computer Science. 2016, 43 (9): 57-60.  doi:10.11896/j.issn.1002-137X.2016.09.010
Abstract PDF(382KB) ( 695 )   
References | Related Articles | Metrics
A novel approach to learn and identify the anomalous behaviors in network was proposed.Unlike previous work,the intrusion detection problem is mapped into the topic model and a classifier is built.Two kinds of connections,namely normal and anomalous ones,are separated before training the model according to the labels of the connections.By analyzing the effect of the parameters,it shows that α (Dirichlet parameter of topics) and the number of topics have positive correlation with the results of prediction,while β (Dirichlet parameter of feature numbers) has negative correlation with the results of prediction.This model was evaluated using KDDCUP’99 dataset.The result suggests that the prediction accuracy is up to 91.69% which outperforms SVM algorithm in normal and anomalous behaviors classification.
Parallelizable Overlapping Community Detection Algorithm Based on Local Expansion
ZHANG Zhong-zheng and LI Jian-wu
Computer Science. 2016, 43 (9): 61-65.  doi:10.11896/j.issn.1002-137X.2016.09.011
Abstract PDF(393KB) ( 530 )   
References | Related Articles | Metrics
An effective way to deal with massive datasets is to decompose an algorithm into a series of irrelevant tasks,and then to execute them in parallel by using open source softwares.Among overlapping community detection algorithms,the methods based on local expansion in its expansion phase only need the information of local communities and their corresponding neighbors,thus they have the possibility to be executed in parallel.In this paper,we proposed a pa-rallelizable algorithm utilizing local expansion for overlapping community detection,and implemented it by using open source software Spark.The algorithm consists of four phases.Firstly,a group of irrelevant central vertices are selected and their corresponding local networks are used as seeds.Secondly,the algorithm filters the selected seeds by removing those whose vertices are weakly connected.Thirdly,the algorithm adopts a batch expansion strategy to expand seeds,by adding a group of neighboring vertices into the local community or removing a group of vertices from the local community.Finally,similar communities are merged.Experimental results based on artificial networks and real world networks show that our method is both accurate and efficient.
Research on Novel Ranking Algorithm of Microblog User’s Influence Based on MapReduce
XU Wen-tao, LIU Feng and ZHU Er-zhou
Computer Science. 2016, 43 (9): 66-70.  doi:10.11896/j.issn.1002-137X.2016.09.012
Abstract PDF(447KB) ( 534 )   
References | Related Articles | Metrics
Featured by instant release,real-time transmission and easy to use,microblog has gradually stepped into the rank of the most popular self-media information platform.User’s influence,which is of great importance to optimize and motivate social information transmission,plays a basic as well as important role in microblog social network.Taking into account the network features of microblog users’ relationship as well as their behaviors, taking Sina microblog as the experimental subject, this paper aimed to introduce the QRank algorithm,a new ranking algorithm based on MapReduce to judge user’s influence.An experiment on the Hadoop platform shows that,with great scalability,QRank algorithm can effectively combine the relationship and behavior features of microblog users and reflect the real influence of users in a more convincing and sufficient way.
Adaptive Parameters Updating Strategy of Context-aware Factorization Machines
YAO Xing, ZHU Fu-xi, YANG Xiao-lan, ZHENG Lin and LIU Shi-chao
Computer Science. 2016, 43 (9): 71-76.  doi:10.11896/j.issn.1002-137X.2016.09.013
Abstract PDF(489KB) ( 520 )   
References | Related Articles | Metrics
Context-aware factorization machine has been successfully applied in the context-aware recommendation system.In the learning algorithm of factorization machines,alternating least-squares is a learning algorithm that fixes other parameters just to find the optimal value of a single parameter,and the number of parameters and the sample size will affect the computational complexity.However,when the number of features is large,the number of parameters increases along with the increase of the number of features,resulting in high computational complexity.Even though some parame-ters have achieved the optimal value,all parameters will be updated in each iteration.This paper mainly improved the para-meters updating strategy of alternating least-squares.Adaptive error index was introduced into the parameter.Updating the parameter or not is co-determined by the weights and the absolute error of parameters,so that each iteration focuses on parameters whose last two iterative values change greatly.This strategy only updates parameters whose adaptive errors are greater than the thresholds.It not only reduces the number of parameters that need to be updated,so as to accelerate the algorithm convergence speed and shorten the operation time,but also the weight of parameters is determined by the error,to correct the error.The results of experiments on Yahoo and Movielens data sets show that the effect of the improved parameter updating strategy is better.
Microblog Text Summarization Based on Entity Relation Network
XUE Zhu-jun, YANG Shu-qiang and SHU Yang-xue
Computer Science. 2016, 43 (9): 77-81.  doi:10.11896/j.issn.1002-137X.2016.09.014
Abstract PDF(1652KB) ( 739 )   
References | Related Articles | Metrics
On the basis of syntax parsing,combining the definition of entity relationship and formalized representation,this paper put forward a method based on directed graph model to reflect the structured relationship between texts,expressing text semantic information,making up for the shortcomings of word frequency characteristics.After that,the corresponding value of each node is measured with improved TPR (Topic-PAGERANK) to represent the importance of the relationship group.Then the corresponding original microblog text of relational tuples is sequentially outputed.Finally,it is proved by experiments that the text summarization extracted by automatic text summarization method based on relational tuple is more comprehensive and less redundant.
Research on Effect of Adding Internal Semantic Relationship into Text Categorization
ZHU Jian-lin, YANG Xiao-ping and PENG Jing-qiao
Computer Science. 2016, 43 (9): 82-86.  doi:10.11896/j.issn.1002-137X.2016.09.015
Abstract PDF(388KB) ( 515 )   
References | Related Articles | Metrics
In order to improve the effect of text categorization on the premise of no addition of the external knowledge,this paper presented a feature matrix-based categorization framework.First,the internal knowledge of corpus is mined and added into the original word-text matrix in different ways.Two common algorithms named SVM and KNN are chosen for contrastive experiment of text categorization in highly territorial legal corpus and domain-wide news corpus.Experi-mental results show that it is generally helpful when adding the semantic relationships extracted from corpus into the original matrix,but the adding method should be chosen according to different classification methods and domain chara-cteristics.
Topological Characterization of AGM Belief Contraction Operator
MENG Hua, YUAN Ya-yan, CHU Jie-lei and WANG Hong-jun
Computer Science. 2016, 43 (9): 87-90.  doi:10.11896/j.issn.1002-137X.2016.09.016
Abstract PDF(321KB) ( 462 )   
References | Related Articles | Metrics
When the background language is finite,there are different semantic methods to characterize belief change ope-rators,which are easy to construct.However,when the background language is infinite,these methods are usually unsuitable.Grdenfors and Makinson proposed a representation model using epistemic entrenchment to characterize belief contraction over an infinite language.But they did not show us how to construct a concrete epistemic entrenchment.In this paper,a new model called “epistemic chain” was introduced to characterize AGM-style belief contraction operators.An epistemic chain was a chain of closed set (about set inclusion) based on a topology on the set of all possible worlds.The relation between epistemic entrenchment and epistemic chain was discussed.Comparing with epistemic entrenchment,epistemic chain is simpler in structure and easier to construct.
Association Rules Mining on Schema-level Interconnected Associated Data
YUAN Liu and ZHANG Long-bo
Computer Science. 2016, 43 (9): 91-98.  doi:10.11896/j.issn.1002-137X.2016.09.017
Abstract PDF(740KB) ( 493 )   
References | Related Articles | Metrics
A schema-level interconnected association rules mining method for large scale associated data was proposed based on the semantic information implied in the associated data set.Instead of mining association rules from separated RDF data sets directly, firstly,we established schema-level linkage between different data sets.The RDF data item pattern generation rules are defined based on the schema-level linked datasets and then the RDF data query techniques are exploited for constructing RDF data items sets.The proposed data item patterns generation rules can extend the data mining objects from a single data set to multi-datasets in the same domain.A Hadoop based implementation plan of association rules mining was designed.The experiment results prove the value of establishing schema-level linkage on linked data and the effectiveness of the proposed method.
Influence Maximization Based on LT+ Model in Social Networks
CAI Guo-yong and PEI Guang-zhan
Computer Science. 2016, 43 (9): 99-102.  doi:10.11896/j.issn.1002-137X.2016.09.018
Abstract PDF(290KB) ( 661 )   
References | Related Articles | Metrics
Influence maximization is a problem of finding a small group of seed nodes in a social network,so that the influence scope of spread in the network is maximized.Kempe and Kleinberg proposed a greedy algorithm which has a wide influence,but its high complexity makes it unsuitable for large social network.Chen and Yuan proposed a heuristic algorithm called local directed acyclic graphs based on linear threshold (LT) model.But LT model only considers the direct influence of neighbors nodes,and ignores the indirect influence between the settled nodes.Therefore,combining with the indirect influence between nodes in the network,we proposed LT+ influence model based on LT model.We also used the local directed acyclic graphs (DAGs) heuristic algorithm to solve the problem of influence maximization,known as LT+DAG algorithm.Extensive experiments were done on real-world dataset to compare the proposed method with other influence maximization algorithms.The result shows that the proposed method can achieve better influence scope and extensibility.
Research on Semi-supervised Learning Based Approach for Lao Part of Speech Tagging
YANG Bei, ZHOU Lan-jiang, YU Zheng-tao and LIU Li-jia
Computer Science. 2016, 43 (9): 103-106.  doi:10.11896/j.issn.1002-137X.2016.09.019
Abstract PDF(323KB) ( 550 )   
References | Related Articles | Metrics
Aiming at the problem of very few corpora resources,a semi-supervised learning based approach for Lao part of speech Tagging was presented.Firstly,a simple probability model is used to obtain a complete dictionary with a small amount of tagged dictionary and untagged corpus,then much more automatically tagged corpus with integer programming are obtained.Finally,a second-order Markov model with sufficient corpus resources is trained to realize a high quality Lao part of speech tagging.This method achieves a good result in Lao part of speech tagging,and its accuracy is up to 89.8%.
URTP:A Factorization Model for Recommendation Based on Users,Regions,Time and Products
HU Ya-hui, YANG Sha, LIU Jing, YU Wei, LI Shi-jun, WANG Jun and FANG Qi-qing
Computer Science. 2016, 43 (9): 107-110.  doi:10.11896/j.issn.1002-137X.2016.09.020
Abstract PDF(419KB) ( 437 )   
References | Related Articles | Metrics
How to recommend a right person with the right products at the right time and place is a challenging topic.First,different users have different culture backgrounds including religion,career,education,preference,etc.Then,different products should be purchased again in different reasonable interval time.And the multi-source heterogeneous,fragmented,various and inconsistent e-commerce data cause problems of sparse data and even cold start.To address these problems,we extracted users’ cultural background with their longitude,latitude and city functional regions.Then,we analyzed the reasonable purchase time and reasonable interval time for different products.And we built a URTP model,which is a factorization model based on users,regions,time and products for recommendation.Experimental results verify the feasibility,effectiveness and efficiency of our algorithm.
Improved Friends Recommendation Algorithm Combining with User Rating Information
TANG Ying, ZHONG Nan-jiang and FAN Jing
Computer Science. 2016, 43 (9): 111-115.  doi:10.11896/j.issn.1002-137X.2016.09.021
Abstract PDF(431KB) ( 527 )   
References | Related Articles | Metrics
Traditional friends recommendation algorithms only consider the topological similarity in calculating the similarity of friends.The similarity of users’ interests is seldomly taken into account,and the recommendation results are often not precise enough.Many existing social networking sites (such as Douban.com) provide the functions of user ratings,i.e. users can give ratings for certain types of items (such as movies).In order to calculate user’s interests similarity,we proposed a method which computes the interests similarity based on the ratings given by users and got more accurate result of recommendation by incorporating the interests similarity into the topological similarity.Firstly,we used cosine similarity to calculate the topological similarity between users.In calculating the interests similarity based on user ratings,we got the users cluster rating similarity matrix through the establishment of a probabilistic model,and derived users’ interests similarity from the rating similarity matrix.Finally,users’ interests similarity and topological similarity were combined to get the final improved friends recommendation algorithm.In order to verify the effectiveness of our method,we applied our method to the crawled user data from Douban website.The experimental results show that our method can effectively improve the accuracy of recommendation results compared with the traditional recom-mendation algorithm based on topology similarity.
SparkDE:A Parallel Version of Differential Evolution Based on Resilient Distributed Datasets Model in Cloud Computing
TAN Xu-jie, DENG Chang-shou, DONG Xiao-gang, YUAN Si-hao, WU Zhi-jian and PENG Hu
Computer Science. 2016, 43 (9): 116-119.  doi:10.11896/j.issn.1002-137X.2016.09.022
Abstract PDF(392KB) ( 571 )   
References | Related Articles | Metrics
MapReduce is a popular cloud computing model which has been applied in data-intensive fields,and Hadoop which is based on MapReduce has been successfully used in dealing with big data.However,when dealing with computation-intensive tasks,particularly iterative computation,frequent loading of Map and Reduce processes will lead to overload.Resilient distributed dataset has been implemented in Spark,and it is an in-memory clustering computing model which can overcome this shortcoming efficiently.In this paper,a parallel version of differential evolution based on RDD (resilient distributed datasets) model named SparkDE was proposed.In SparkDE,the whole population is divided into several islands which evolve on their own,and then each island is deployed into a partition of RDD.After evolution for predefined generation in each island,migration operator is used calculation between islands.A wide range of benchmark problems are adopted to conduct numerical experiments.Compared with MapReduce (MRDE) based DE and classical DE,the results show that SparkDE can achieve higher accuracy of solution and faster speed of computation.The speedup of SparkDE is obvious.Thus SparkDE can serve as the next generation of optimizer in cloud computing.
Efficient and Effective Clustering Algorithm for Asynchronous Time Series of Clinical Laboratory Indicators
CHEN De-hua, HAN Xue-shi, LE Jia-jin and ZHU Li-feng
Computer Science. 2016, 43 (9): 120-123.  doi:10.11896/j.issn.1002-137X.2016.09.023
Abstract PDF(411KB) ( 640 )   
References | Related Articles | Metrics
Clustering for asynchronous time series of clinical laboratory indicators,and finding the patient group with similar variation trends of clinical laboratory indicators,have a very important value for the conduct of precision medicine.Taking into account the frequency of inspection and the testing time points of different patients are not fully synchronized, asynchronous time series were preprocessed to achieve the synchronization of different time dimensions and time points.On this basis,we improved the DBScan algorithm by introducing a user-defined parameter namely noises share NoisePro.Then,we proposed a LabTS-CLU time series clustering algorithm of asynchronous clinical test indicators based on density divided thoughts.Finally,experimental results on the time series of glycated hemoglobin dataset of more than 100 thousand diabetics in the past 10 years from a hospital demonstrate the effectiveness of the proposed algorithm.
Modeling and Analyzing of WSNs Data Gathering Protocol Based on UPPAAL
FENG Ya-chao, YANG Hong-li, WANG Fei, WU Wen-jia and QIN Sheng-chao
Computer Science. 2016, 43 (9): 124-130.  doi:10.11896/j.issn.1002-137X.2016.09.024
Abstract PDF(2582KB) ( 510 )   
References | Related Articles | Metrics
Wireless Sensor Networks is widely used in various types of data gathering systems,such as residential wireless meter reading (including water,electricity and gas meters) systems.The correctness and rationality of the designing of data gathering protocol are the key factors affecting the normal operation of the network.We proposed the method of modeling and analyzing of WSNs data gathering protocol based on UPPAAL towards the demand of real-time feature.Considering the input model of UPPAAL is more complex than the general terms of automata model,we established the general terms of time automata model of data gathering protocol at first,and then transfered it to the input model of UPPAAL.The effectiveness of the method is illustrated by modeling and analyzing for an actual wireless meter reading data gathering protocol WM2RP.The result shows some properties which are related to the safety and reliability can be satisfied on the protocol.The exception model and energy consumption model of WM2RP are further established to analyze the protocol from the multiple angle.
Research of Cache Management Strategy Based on Splay Tree
YAO Zhi-hai, XU Hong-zhe, LI Wen and WU Xia
Computer Science. 2016, 43 (9): 131-134.  doi:10.11896/j.issn.1002-137X.2016.09.025
Abstract PDF(345KB) ( 505 )   
References | Related Articles | Metrics
Taking the enterprise’s internal network storage application as background,we put forward a cache management strategy based on splay tree to organize buffer memory space.We improved the structure and operation of splay tree and took it as the index structure for organizing and managing cached data.Finally,according to the actual application requirements,we analyzed and designed the model of cache management and made it come true.The experimental results show that,through combining with the improved splay tree algorithm,the buffer memory space and cached data can be effectively organized and managed,therefore,the cache management strategy proposed in this paper has good real time property and the efficiency of accessing data can also be improved.
Key of Internet Modeling-Looking Inside Inter-AS Router-level Connection
LI He-shuai, ZHU Jun-hu, WANG Qing-xian, QIU Han and ZHOU Tian-yang
Computer Science. 2016, 43 (9): 135-139.  doi:10.11896/j.issn.1002-137X.2016.09.026
Abstract PDF(408KB) ( 658 )   
References | Related Articles | Metrics
Currently,the AS-router double layer modeling is the optimal scheme for constructing Internet-scale router-level topology model,while it needs researchers to understand the characteristics of the Inter-AS router-level network topology.This paper pointed out that the type of AS has a great influence on Inter-AS router-level connection.After defining the mappings of Inter-AS parameters,we analysed the phenomenolog on CAIDA network data and obtained the distribution function of different Inter-AS parameters.And then we solved the problem of the combination of AS-level topology and Inter-AS router-level topology.At last,more problems of the Internet modeling that need to be solved were discussed.
Multi-stage Feedback Fountain Code for Deep Space Transmission Scheme
CHEN Kang-ni, QIAN Li-ping and CHEN Qing-zhang
Computer Science. 2016, 43 (9): 140-145.  doi:10.11896/j.issn.1002-137X.2016.09.027
Abstract PDF(527KB) ( 533 )   
References | Related Articles | Metrics
A transmission scheme of deep space communication based on multi-stage fountain code was proposed to cope with the high error rate and long propagation delay in deep space communications.In particular,the framework of co-ding and transmission were first presented,and then the coding efficiency and transmission time of this scheme were theo-retically evaluated comparing to the present transmission scheme without feedback.Additionally,the performance of these two schemes and the decoding forward transmission scheme at difference bit error rates and transmission distances were analyzed.The results show that the proposed transmission scheme improveed the coding efficiency and reduces the file propagation delay.
Evolutionary Game Theory-based Access Control Study for P-persistent CSMA Networks
WANG Le, MAO Jian-lin, ZHU Hao-fu and GUO Ning
Computer Science. 2016, 43 (9): 146-151.  doi:10.11896/j.issn.1002-137X.2016.09.028
Abstract PDF(531KB) ( 498 )   
References | Related Articles | Metrics
Considering the noncooperation system behavior of p-persistent CSMA network where the wireless channel error may occur,this paper established a p-persistent CSMA evolutionary game model (EGP-CSMA).In this model,the unique evolutionary stable strategy (ESS) is derived and the optimal ESS to maximize saturation throughput and minimize average energy consumption are calculated respectively.Then,the effect of payoff delay,success payoff and bit error rate are studied on the convergence of evolutionary dynamics to the optimal ESS.Numerical simulation results show that,as the bit error rate is constant and the payoff delay is short,a proper cost and payoff can make the multiple access game reach an evolutionary stable status at the optimal transmission probability.Thus,the p-persistent CSMA network can be stable and perform optimally.
Opportunistic Routing Algorithm Based on Partial Network Coding for Wireless Networks
WANG Zhen-chao, CAI Zhi-jie and XUE Wen-ling
Computer Science. 2016, 43 (9): 152-155.  doi:10.11896/j.issn.1002-137X.2016.09.029
Abstract PDF(416KB) ( 507 )   
References | Related Articles | Metrics
A new opportunistic routing algorithm for wireless network based on partial network coding (ORAPNC) was proposed,which combines the advantages of opportunistic routing and network coding.In order to avoid the bifurcation transmission of data packets and benefit the implementation of the coordination mechanism among forwarding nodes,firstly,ORAPNC establishes a fixed path using expected transmission count as path metric,meanwhile gathers the candidate forwarding nodes in the vicinity of this fixed path.Then,ORAPNC adopts a new forwarding nodes coordination mechanism (FNCM) to achieve per-hop packet transmission for the sake of reducing redundant data packets in the network sufficiently.Simulation results show that,comparing to other routing protocols,ORAPNC performs more effectively on improving network throughput,decreasing average delay of decoding the original data packets at destination node.
Performance Model for Network-coding-aware Opportunistic Routing in Wireless Networks
TAO Wen, JIN Ling, BAI Guang-wei and SHEN Hang
Computer Science. 2016, 43 (9): 156-159.  doi:10.11896/j.issn.1002-137X.2016.09.030
Abstract PDF(316KB) ( 376 )   
References | Related Articles | Metrics
This paper proposed a performance model for network-coding-aware opportunistic routing in wireless networks working together with multi-priority ACK response strategy,where the focus is on exploring the impact of the packet loss of wireless channel on the network’s performance as relay nodes increases.A Markov chain model is deve-loped to characterize DCF random channel access mechanism,thus the important parameters with respect to network performance are obtained.The mathematical results show that the joint design of opportunistic routing and network coding can greatly improve the packet delivery ratio and end-to-end throughput when increasing the number of relays.
Modeling and Analysis on Facilities Collaboration Network
LI Wen-xiang, WEI Xia, JIANG Hao and SHENG Yu-xia
Computer Science. 2016, 43 (9): 160-164.  doi:10.11896/j.issn.1002-137X.2016.09.031
Abstract PDF(1051KB) ( 414 )   
References | Related Articles | Metrics
There exist complex collaboration relations among production resources in regional enterprise cluster.To achieve efficient cloud manufacturing services,this paper explored how to mine and utilize these relations for controlling the collaboration behaviors.It described the collaboration relations based on generalized social collaboration network,and proposed the method for building Facility Collaboration Network (FCN).It further designed the dynamic growth process of FCN for different facility selection strategies (random,balanced,random with preference and balanced with preference).Based on the metrics such as network scale,degree distribution and average shortest distance,it analyzed the statistical characteristics of an instance of FCN,and stated that balanced strategies possess larger network scale,while random strategies exhibit obvious scale-free and small-world characteristics.At last,it proposed strategies (dynamic weighing of facilities and concentrative processing of successive subtasks) for employing facilities efficiently.
Research on Access Control Model in Multi-clouds Storage System Based on CP-ABE
YIN Kai-ze and WANG Hai-hang
Computer Science. 2016, 43 (9): 165-168.  doi:10.11896/j.issn.1002-137X.2016.09.032
Abstract PDF(1720KB) ( 459 )   
References | Related Articles | Metrics
Applying the access control system based on ciphertext-policy attributes-based encryption (CP-ABE) in single cloud to multi-clouds storage system will encounter a problem of policy conflict.Thus,an attributes mapping scheme was designed and an access control model in multi-clouds storage system based on CP-ABE was provided.The attributes mapping scheme was designed based on the access construction tree of CP-ABE and the types of attributes’ value that is supported.At last,the framework of this model and its workflow were elaborated.The effectiveness of the model is verified by building a simple prototype system,and the performance of the prototype system is analyzed.The proposed model has theoretical value and actual meaning for the research of access control in multi-clouds storage system.
Military Aeronautical Communication Spectrum Sharing Trust Mechanism Based on Cloud Model
XU Xue-fei, LI Jian-hua, YANG Ying-hui and GUO Rong
Computer Science. 2016, 43 (9): 169-174.  doi:10.11896/j.issn.1002-137X.2016.09.033
Abstract PDF(487KB) ( 517 )   
References | Related Articles | Metrics
In response to the particularity of military aeronautical communication spectum sharing,a spectrum sharing trust mechanism based on cloud model was proposed.By the use of trust relevant theories and the introduction of the cloud model theory,a system framework of spectrum sharing based on the cloud model theory was constructed,and the mutual relationship of spectrum sharing system was confirmed.The clear idea and concrete method of trust evaluation based on the cloud model was proposed at last.The trust evaluation approach can implement the conversion from the fuzzy qualitative trust attributes to the precise quantification trust levels,which could resolve the series problems effectively during the establishment of spectrum sharing trust relationships.The trust evaluation experimentation was designed.The result show that the trust evaluation model is of validity and practicability,which provide the effective support of spectrum sharing strategy based on the trust mechanism.
Cooperative Address Knocking Based Covert Authentication
CAO Xu, ZHU Yue-fei and FEI Jin-long
Computer Science. 2016, 43 (9): 175-179.  doi:10.11896/j.issn.1002-137X.2016.09.034
Abstract PDF(428KB) ( 462 )   
References | Related Articles | Metrics
With the development of cloud computing,it is inevitable that many security problems arise.Unauthorized service access is one of the most important threats.Based on the new features of IPv6 address,we proposed a new network security technique called cooperative address knocking,which can be seen as an undetectable authentication.It is a form of host-to-host communication which relies on deliberate communication attempts from some cooperative nodes.These connection attempts are monitored by a daemon which interprets the interface identifier of destination IP addresses as information.The theoretical and empirical analysis demonstrate that CAKCA scheme can effectively conduct undetectable authentication and prevent the exposure of existence of the important host.The theoretical analysis and simulation results show that the proposed scheme can effectively improve the level of network security.
High Efficiency Multi-authority Cloud Access Control Scheme
ZHOU Peng-xu and LI Cheng-hai
Computer Science. 2016, 43 (9): 180-183.  doi:10.11896/j.issn.1002-137X.2016.09.035
Abstract PDF(398KB) ( 458 )   
References | Related Articles | Metrics
For solving the overhead problems of users in the multi-authority access control schemes,a HE-MA-ACS scheme was proposed.Outsourced decryption is introduced based on the hierarchical authorization structure,so large part of the decryption overhead is moved to the CSP.Furthermore,fine-grained attribute revocation is achieved and the users can not participate in the operation when their attributes are revoked.The correctness,security,calculated and storage performance were analyzed.Experimental results demonstrate the superiority of overhead in user storage,access communication,decryption and the computation costs when attribute is revoked as well.The scheme effectively reduces the burden on the user side and improves the efficiency of decryption.
Privacy Protection from Implicit Attacks in Cloud Computing Environment
TAO Lin-bo, SHEN Jian-jing, XUE Meng and CAI Li-gang
Computer Science. 2016, 43 (9): 184-187.  doi:10.11896/j.issn.1002-137X.2016.09.036
Abstract PDF(1159KB) ( 418 )   
References | Related Articles | Metrics
Privacy protection in cloud computing environment has become the key to the rapid development of cloud computing,until now more researches focus on the explicit protection methods of privacy protection such as encryption,lacking of enough attention to implicit attack,such as data mining attack.A protection model was established in this paper according to how to protect privacy in cloud computing environment against data mining attack.The model selects several key objectives as the key protection during the life cycle of user’s data,including input,storage,reduction and destruction.Then correlative modules were established according to the key objectives corresponding to data privacy.Finally,the effectiveness of this model was proved through the complexity comparison.
Research and Implementation of EFI OS Loader Security Reinforcement Technology
WU Wei-min, CHEN Dong-xin, LAI Wen-xin and SU Qing
Computer Science. 2016, 43 (9): 188-191.  doi:10.11896/j.issn.1002-137X.2016.09.037
Abstract PDF(1908KB) ( 1017 )   
References | Related Articles | Metrics
By analyzing the safety of architecture and boot procedure of unified extensible firmware interface (UEFI),it is found that the credibility verification of EFI OS Loader has security risks,which can lead to the hijack of Windows startup process.To avoid the security risks,considering from the three layers of file isolation protection,boot authentication and system critical region protection,a three-layer security reinforcement plan based on USB Key,the dynamic password cell phone token and EFI antivirus software was proposed.Storing the EFI OS Loader file in the USB Key and encrypting it can achieve the file protection.The dynamic password authentication server is placed in the USB Key,and the combination of both mechanism can achieve a high intensity boot authentication.Designing and developing an EFI application security software following the UEFI specification can achieve the protection of the key region of system.The results show that the dual authentication and security mechanism of the program make up the relevant security vulnerabilities,and enhance the security of computer systems during startup.
Approach of Android Applications Intent Injection Vulnerability Detection Based on Static Taint Analysis
WANG Yun-chao, WEI Qiang and WU Ze-hui
Computer Science. 2016, 43 (9): 192-196.  doi:10.11896/j.issn.1002-137X.2016.09.038
Abstract PDF(435KB) ( 1198 )   
References | Related Articles | Metrics
As a message carrier in the process of component communication of Android application,Intent can be malformed by an attacker,leading to security risk of malicious component injection.A detection approach based on static taint analysis was presented.On the basis of building call graph and control flow graph of Android application,by trackingthe taint propagation with in and between components,the potential Intent injection vulnerability can be detected.This method is used to test four types of benchmark and fifty third-party applications,and the experimental results show the feasibility and effectiveness of the proposed approach.
Distributed Parallel Semantic Coding Algorithm for RDF Data
ZHENG Cui-chun and WANG Jing-bin
Computer Science. 2016, 43 (9): 197-202.  doi:10.11896/j.issn.1002-137X.2016.09.039
Abstract PDF(556KB) ( 575 )   
References | Related Articles | Metrics
The existing distributed parallel compression coding algorithms for RDF data do not consider combining with the ontology file,resulting in encoded RDF data without any semantic information,which is not conducive to the distribu-ted query or reasoning. To solve these problems,a method named SCOM (Semantic Code with Ontology on MapReduce) was proposed to complete the semantic parallel coding for RDF data.Firstly,the algorithm combines the ontology of RDF data to build the class and attribute relationship model.The triple items are encoded and a dictionary table is generated after classifying and filtering triples.Finally,the coding for RDF data with semantic information and regularities is completed.In addition,SCOM algorithm can easily revert the encoded RDF data file to their original file.Experimental results show that SCOM algorithm can achieve the parallel coding of large-scale data efficiently.
Partially Regenerating Codes Combined with Replicas
DING Bing-chen and LI Wei-zhong
Computer Science. 2016, 43 (9): 203-208.  doi:10.11896/j.issn.1002-137X.2016.09.040
Abstract PDF(441KB) ( 409 )   
References | Related Articles | Metrics
n,k,d) Regenerating codes(RC) significantly reduce repair bandwidth by allowing storage nodes to send the linear combinations of their data to the newcomer and increasing repair degree d.But they bring in more participating nodes and disk I/O.To solve this problem,this paper introduced partially regenerating codes(PRC) which combine RC (n,k,d,λ,θ) with replicas.PRC have also a threshold function and two special points:minimum-storage point and minimum-bandwidth point.PRC can simultaneously reduce repair bandwidth and disk I/O by utilizing repair degree d and replica factor θ.When the storage capacity of all nodes are the same,the repair bandwidth and disk I/O per node of PRC are superior to that of RC.The results of quantitative comparison also show that,comparing to RC,on minimum-storage point,PRC have less mean repair bandwidth and disk I/O;on minimum-bandwidth point,PRC have less mean disk I/O and mean repair bandwidth to similar RC.What’s more important is that PRC is achievable when d≤n-2.
Clustering Ensemble Algorithm for Large-scale Mixed Data Based on Sampling
PANG Tian-jie and LIANG Ji-ye
Computer Science. 2016, 43 (9): 209-212.  doi:10.11896/j.issn.1002-137X.2016.09.041
Abstract PDF(309KB) ( 469 )   
References | Related Articles | Metrics
In clustering analysis,one of the important problems is mixed data clustering.The clustering of existing algorithms is mainly based on similarity measurement of all samples.Therefore,the efficiency of clustering for large-scale data is not high.So we designed a new sampling strategy and proposed an ensemble algorithm for large-scale mixed data based on sampling.This new algorithm clusters subsets which are obtained by the use of the new sampling strategy respectively and the final clustering results can be gotten by clustering ensemble.Experiment shows that the efficiency of algorithm is improved significantly and the clustering validity indexes are almost the same compared with the modified K-prototypes algorithm.
Study of Scientific Paper Recommendation Method Based on Unified Probabilistic Matrix Factorization in Scientific Social Networks
WU Liao-yuan, JIANG Jun and WANG Gang
Computer Science. 2016, 43 (9): 213-217.  doi:10.11896/j.issn.1002-137X.2016.09.042
Abstract PDF(416KB) ( 458 )   
References | Related Articles | Metrics
In recent years,the number of scientific papers in scientific social networks has grown at an explosive rate.It is difficult for researchers to find scientific papers related to their research.Therefore,the paper recommendation for researchers was proposed to solve this problem.However,many problems exist in traditional paper recommendation methods,especially for the fact that a lot of social information in scientific social network are not fully used,resulting in poor quality of paper recommendation.Therefore,this research proposed a new paper recommendation method for researchers in scientific social networks based on the unified probability matrix factorization.This method incorporates social tag information and group information into traditional matrix factorization.In order to verify the validity of the proposed method,we crawled data from a famous scientific social network,i.e.CiteULike,to conduct experiments.Experimental results show that the proposed method gets the best recommendation results at the two evaluation metrics,i.e. Precision and Recall,compared to other traditional recommendation methods.The proposed method is linear with respect to the number of observed data,and performs well in scalability.
Improved Particle Filter and its Application in GPS/DR Integrated Positioning System
DU Hang-yuan, WANG Wen-jian and BAI Liang
Computer Science. 2016, 43 (9): 218-222.  doi:10.11896/j.issn.1002-137X.2016.09.043
Abstract PDF(412KB) ( 555 )   
References | Related Articles | Metrics
An improved particle filtering algorithm based on the ensemble Kalman filter (EnKF) was proposed in this paper starting with the selection of importance density function of the particle filter.At each time instant,the importance density function is generated by EnKF which fuses the latest observation information and propagates the system states by using a collection of sampled state vectors,called an ensemble.In this way,the importance density function can be very close to the true posterior probability.Furthermore,to avoid the particle impoverishment problem,the Markov Chain Monte Carlo method was introduced after resampling process.In the simulation,the developed filter was compared with standard particle filter,extended Kalman particle filter and unscented particle filter in GPS/DR integrated system.The simulation results demonstrate the validity of the developed algorithm.Under the same conditions,the new filter is superior to other particle filtering algorithms with the respect to estimation accuracy,as well as it controls the computational load effectively.It is also found that the new filter can obtain outstanding performance even with a small number of particles.
CNN Training Algorithm Based on Co-studying of Multiple Classifiers
CHEN Wen, ZHANG En-yang and ZHAO Yong
Computer Science. 2016, 43 (9): 223-226.  doi:10.11896/j.issn.1002-137X.2016.09.044
Abstract PDF(1014KB) ( 643 )   
References | Related Articles | Metrics
Convolutional neural network (CNN) is a kind of important deep neural network.However,amounts of labeled samples are needed in the training process of CNN,which largely limits its applications.For the problem,the co-training process of CNN was analyzed,and a co-training algorithm CAMC based on the iterative evolution of classifiers was given.The advantages of CNN and co-training of multiple classifiers are combined in CAMC.Firstly,multiple features are drawn using different convolutional kernels to build up different CNN classifiers.Then to continually improve the classification performance,some labeled samples together with many unlabeled samples are employed to co-train the multi classifiers.The experimental results based on the standard data sets of human expression demonstrates that,compared with traditional expression recognition methods LBP and Gabor,the performance of CAMC can be continually improved,thus it has higher classification accuracy.
Collaborative Filtering Algorithm Integrating Multiple User Behaviors
GAO Shan, LIU Wei, CUI Yong, ZHANG Qian and WANG Zong-min
Computer Science. 2016, 43 (9): 227-231.  doi:10.11896/j.issn.1002-137X.2016.09.045
Abstract PDF(414KB) ( 474 )   
References | Related Articles | Metrics
Collaborative filtering is one of the most successful techniques among personalized recommender systems,and is widely used in the field of e-commerce,social networks etc.Due to the complexity and diversity of the personal health behaviors,it causes low accuracy of the recommendation algorithms when applied to personalized medicine recommendation.To deal with this problem,a new collaborative filtering algorithm integrating multiple user behaviors was proposed.The weighting factor and the overlap-dependency are introduced in classical similarity computing,and they can measure the effects of different user behaviors on the recommendation quality.Experiments on the Top-md dataset show that the new algorithm can fully express the user’s preferences and wishes for medical treatment,and can effectively improve the quality of personalized medicine recommender systems.
Variable Precision Rough Set Model Based on Extended Dominance Relations
LI Yan, JIN Yong-fei and MA Hong-yan
Computer Science. 2016, 43 (9): 232-237.  doi:10.11896/j.issn.1002-137X.2016.09.046
Abstract PDF(453KB) ( 391 )   
References | Related Articles | Metrics
The variable precision rough set (VPRS) model based on dominance relations extends equivalence relations in traditional rough sets to dominance relations,and combines with the idea of variable precision to define the relevant concepts.Therefore,it can deal with preference-ordered information with certain fault tolerance degree.However,the definition of traditional dominance relation is still too strict,in which object x is superior to object y only when all attribute values of x are superior to that of y.This definition is difficult to be satisfied especially when the number of attributes is large.This will lead to smaller dominance,and even worse,it will affect the extraction of decision rules and then the process of decision making.To address this problem,the concept of dominance relation was extended by introducing a parameter and then the dominance set and approximation sets were correspondingly defined based on this extended do-minance relation.The extended VPRS model was also developed,and the coverage rate and the test accuracy were used as evaluation criteria to for model.Finally,an illustrative example was given and the experimens on UCI data were conducted to compare the proposed extended model with the traditional VPRS model.
Gauss Cloud Model Based on Uniform Distribution
CHEN Hao and LI Bing
Computer Science. 2016, 43 (9): 238-241.  doi:10.11896/j.issn.1002-137X.2016.09.047
Abstract PDF(421KB) ( 669 )   
References | Related Articles | Metrics
The cloud model has established the uncertain transformation between qualitative concepts and their quantitative description.The universality of Gauss distribution and Gauss membership function prepares the ground for the universality of Gauss cloud model,and up to now the theory of cloud model and its applied researches have mainly focused on Gauss cloud model.In this article the uniform distribution was introduced into cloud model and Gauss cloud model was extended,thus resulting in two models:uniformly distributed cloud model and uniform Gauss cloud model which offer a new way for uncertainty description.Finally,the multi-dimensional uniform Gauss cloud model was used to simulate the growth process of fractal tree.The result indicates that the cloud model can be used to effectively simulate the fractal phenomena in nature.
Credit Evaluation of P2P Lending User Based on Granular Computing and Information Fusion
ZHAO Ying-xiu, LIU Wen-qi, LI Jin-hai and ZHAO Ning
Computer Science. 2016, 43 (9): 242-246.  doi:10.11896/j.issn.1002-137X.2016.09.048
Abstract PDF(411KB) ( 414 )   
References | Related Articles | Metrics
P2P lending faces enormous credit risk when it is explosively growing .Firstly,we put forward a credit evalua-tion method namely comprehensive evaluation method and varified it by using 360 data.Secondly,we proposed information fusion method using granular computing to make third-party data information fusion with existing credit evaluation values.This method not only adds on the original credit evaluation value information,but also makes each other getting verified.At last,we verified information fusion method using granular computing through simulating the third party data.
Improvement of Lucene Sorting Algorithm Fusing Location-related and Probabilistic Sorting
HU Bo and JIANG Zong-li
Computer Science. 2016, 43 (9): 247-249.  doi:10.11896/j.issn.1002-137X.2016.09.049
Abstract PDF(336KB) ( 593 )   
References | Related Articles | Metrics
Sorting document retrieval results and text classification technology is the core technology to solve vertical search,personalized information retrieval,information filtering and other related issues.In order to improve the performan-ce of retrieval systems,an improved method for integrating location-related and probabilistic sorting was proposed for Lucene default sorting algorithm.Taking into account the document relevance impact of query’s location information and probabilistic sorting,the scoring formula of Lucene default sorting algorithm is improved using the probability value of document relevance based on naive Bayesian classification algorithm and the weights of location-related query.Experimental results show that this improvement can effectively improve the accuracy of vertical search,allowing users to have better vertical search experience.
Improved Intuitionistic Fuzzy Genetic Algorithm for Nonlinear Programming Problems
MEI Hai-tao, HUA Ji-xue and WANG Yi
Computer Science. 2016, 43 (9): 250-254.  doi:10.11896/j.issn.1002-137X.2016.09.050
Abstract PDF(363KB) ( 690 )   
References | Related Articles | Metrics
To solve the multidimensional nonlinear programming problem,an improved intuitionistic fuzzy genetic algorithm(IFGA) was proposed.The membership and nonmembership degrees of individuals are defined by the individual fitness of the genetic algorithm in each iteratine optimization,and the problem is transformed to the intuitionistic fuzzy nonlinear programming problem to adjust crossover and mutation rates.And the paper proposed an improved selection operator.The individuals are divided into four same size groups,and the group with poor fitness is selected and copied randomly to increase the diversity and competitiveness,because the implicit optimization information of non-feasible solution is reserved.The simulation results indicate the IFGA is feasible and effective.
Mobile Social Recommendation Based on Unified Probabilistic Matrix Factorization
XIONG Li-rong, LIU Jian and TANG Ying
Computer Science. 2016, 43 (9): 255-260.  doi:10.11896/j.issn.1002-137X.2016.09.051
Abstract PDF(583KB) ( 464 )   
References | Related Articles | Metrics
It has become the main task of mobile recommender systems to further improve the prediction quality and solve the data sparsity and cold-start problems that may exist by employing mobile context and mobile social network information etc.We combined users,services and users’ social network information for recommendation to alleviate the data sparsity and cold-start problems by using the factor analysis method based on matrix factorization (MF).In order to increase the trust matrix density,in this paper we imported the indirect trust relationship,and then proposed a trust relationship calculation method which only use the mobile social network information to build trust matrix to reduce the user’s active identification for trust relationship.And the trust calculation method is in line with the characteristics of mobile social network.The experimental results show that the introduction of the indirect trust relationship can improve the prediction accuracy,and our method outperforms some existing MF methods and traditional collaborative filtering algorithm in the aspect of accuracy,especially in the circumstance that users have made very few ratings or even none at all.
Extract Summarization Method Based on Lex-PageRank in Chinese Microblog
ZHU Ming-feng, YE Shi-ren and YE Ren-ming
Computer Science. 2016, 43 (9): 261-265.  doi:10.11896/j.issn.1002-137X.2016.09.052
Abstract PDF(430KB) ( 448 )   
References | Related Articles | Metrics
In recent years,since the rise of personal-media caused a huge public opinion crisis,how to discover hot topics from the fragmentation of the mass microblogging information and extract useful information has become a major challenge.In this background,we proposed an extract summarization method based on improved Lex-PageRank algorithm.In this program,we make a simple clustering and use these clustering results as experimental data.With due consideration for the time characteristics of microblog influence cycle and weight attribute,the improved Lex-PageRank algorithm is combined with MMR algorithm to get many texts from clustering results to generate a summary.The experiment based on Sina Weibo indicates that our method can extract critical information effectively from mass texts.
Method of Personalized Collaboration Filter Recommendation Based on Bayesian Network
FU Yong-ping and QIU Yu-hui
Computer Science. 2016, 43 (9): 266-268.  doi:10.11896/j.issn.1002-137X.2016.09.053
Abstract PDF(255KB) ( 634 )   
References | Related Articles | Metrics
Lacking of high efficiency personal recommendation in recommendation system,we proposed a new method in collaboration filtering recommendation system by using semantic checking.We check the result of collaboration filtering based on user item by Bayesian semantic to eliminate the item of lower probability,and to select the higher probability item to users.In constructing the Bayesian semantic check network,we add an emotion field named “fancy” by questionnaire survey and information feedback,and we decide user’s emotion for some goods to ensure the scientific of semantic checking network.Experiments show that the method can eliminate the items with low user preferences and improve the satisfaction degree of the users.
Factorization Machine Based on Differential Evolution
YU Fei, ZHAO Zhi-yong and WEI Bo
Computer Science. 2016, 43 (9): 269-273.  doi:10.11896/j.issn.1002-137X.2016.09.054
Abstract PDF(399KB) ( 516 )   
References | Related Articles | Metrics
Factorization machine(FM) is a new machine learning algorithm based on the matrix factorization.It can be used to deal with the regression problems,classification problems and ranking problems.The solution of parameters in this model is based on the optimization method of gradient.However,under the condition of small amount of samples,the optimization method based on gradient has a slow convergence rate and may stick into local optimum.Differential evolution(DE) is a heuristic global optimization algorithm.It has a fast convergence rate.In order to improve the accuracy of FM,we proposed the DE-FM algorithm,which searches the best parameters of FM model with DE algorithm.We compared DE-FM with FM on the Diabetes dataset,the Horse-Colic dataset and the Music dataset,and the result shows that DE-FM can improve the accuracy.
Research on Video Copy Detection Algorithm Based on Spatial-Temporal Domain Informative Fusion
YAN Cong, JI Mo-xuan and JI Qing-ge
Computer Science. 2016, 43 (9): 274-279.  doi:10.11896/j.issn.1002-137X.2016.09.055
Abstract PDF(1305KB) ( 544 )   
References | Related Articles | Metrics
In order to effectively utilize unique spatial-temporal domain characteristics of video to enhance the robustness and accuracy of copy detection algorithm,this paper proposed a fast video copy detection algorithm based on spatial-temporal domain informative fusion,which includes a fingerprint extraction algorithm based on spatial-temporal domain informative fusion and two kinds of matching search algorithms of which one is based on inverted-file index and the other is based on matching state machine with asynchronous window strategy.The fingerprint extraction algorithm firstly forms the spatial-temporal domain informative frame by video segmentation,then partitions the informative frame into blocks and extracts DCT coefficient with its median value as threshold to obtain video fingerprint.The matching search algorithm based on inverted-file index sets up inverted-file index table by binary characteristics of fingerprint,then quickly queries fingerprint according to the index table.Combining the matching search algorithm based on matching state machine with asynchronous window strategy,we can change search scope and step size by matching state with nearest neighbor.Meantime,asynchronous window strategy can adopt different extract strategies in online and offline process to accelerate the whole search.The experimental results show that our fingerprint extraction algorithm is robust in the case of Gaussian noise,adding subtitles,spatial shift,rotation and frame drop,and the proposed schemes tend to have great improvement in time efficiency.
Block-coded Video Deblocking Based on Low-rank Tensor Recovery
CHEN Dai-bin and YANG Xiao-mei
Computer Science. 2016, 43 (9): 280-283.  doi:10.11896/j.issn.1002-137X.2016.09.056
Abstract PDF(1528KB) ( 507 )   
References | Related Articles | Metrics
Block-coded videos suffer from the blocking artifacts after being decoded.In order to solve this problem,a block-based deblocking method using low-rank tensor recovery was proposed.First,three order tensor is constructed through clustering similar blocks in video sequence.Then,according to low-rank property of background tensor and sparsity of blocking artifacts,the proposed approach utilizes the augmented Lagrange multiplier method which extends to tensor to solve the low-rank tensor recovery problem.The proposed approach utilizes tensor model to preserve the structural properties of high dimensional data.Experimental results show that it can obtain higher PSNR value and better visual effect comparing with traditional deblocking methods.
Multiplayer Online Virtual Experiment System Based on Kinect Somatosensory Interaction
SUN Bo-wen, ZHANG Jia-liang, CAI Ya-fei and GUO Wen-lan
Computer Science. 2016, 43 (9): 284-288.  doi:10.11896/j.issn.1002-137X.2016.09.057
Abstract PDF(1613KB) ( 1132 )   
References | Related Articles | Metrics
In order to meet the needs of allowing more than one person conducting a certain virtual experiment at the same time in different places,we proposed a multiplayer online virtual experiment system built by Kinect somatosensory equipment and Unity 3D engine.In this system,the virtual experiment scenes are built by Unity 3D engine and experimental equipments are made by using 3D Max which is well-known as a modeling software.The multi-player online function is completed by the network communication technology.To make users feel real,the body gesture caught by Kinect somatosensory technology can be used to control first-person-character to walk around,grab experimental equipments and manipulate the menu in the virtual scene.The practical result shows that the Kinect gesture recognition has good accuracy and robustness,and is hardly affected by light situations and complex background.At the same time,the communication between the server and client is stable enough to build a remote virtual experiment system,and the system can achieve a high immersion with lower cost.
Unsupervised Feature Learning Based Interest Point Detection Algorithm
ZHOU Lai-en and WANG Xiao-dan
Computer Science. 2016, 43 (9): 289-294.  doi:10.11896/j.issn.1002-137X.2016.09.058
Abstract PDF(2912KB) ( 580 )   
References | Related Articles | Metrics
Interest point is of great importance in digital image processing as a kind of critical feature at low level.So the interest point detection is the committed step in image registration,image retrieval and image recognition.In this paper,an unsupervised feature learning based interest point detection (UFL-ID) was presented based on the fact that interest points have stronger feature convolution response than others.The new UFL-based interest point detection algorithm firstly learns low level features in digital images,evaluates the information content and isotropy of learned features,and finally uses features and its evaluation to find interest points.The comparison result demonstrates that using UFL produces great improvements of repeatability and anti-noise property.
Gait Recognition Based on Decomposition of Optical Flow Components
LUO Zheng-ping, LIU Yan-jun and YANG Tian-qi
Computer Science. 2016, 43 (9): 295-300.  doi:10.11896/j.issn.1002-137X.2016.09.059
Abstract PDF(1517KB) ( 873 )   
References | Related Articles | Metrics
Gait recognition has gained tremendous attention for its characteristics of long distance and hard-to-disguise.Aiming at the problem of insufficient information of the existing feature extraction method,this paper proposed a novel gait recognition method based on the decomposition of optical flow components.The positive transverse or longitudinal components in gait optical flow image are decomposed by rows and columns,then the transverse and longitudinal components of optical flow for each row or colum are calculated,and four feature vectors are obtained.The four vectors are fused according to their weight in recognition.Principle components analysis and linear discriminant analysis techniques-are combined,and dynamic time warping algorithm is used to match.Finally,K-nearest neighbor algorithm is used for classification.Experiments on CASIA Database B and C show that,the proposed method achieves recognition accuracy of 97%,90% and 64% respectively under the conditions of normal,backpack-wearing and coat-wearing,88% and 87% under conditions of slow walking and fast walking.
Parameter-less Supervised Kernel Locality Preserving Projection and Face Recognition
GONG Qu and XU Kai-qiang
Computer Science. 2016, 43 (9): 301-304.  doi:10.11896/j.issn.1002-137X.2016.09.060
Abstract PDF(394KB) ( 445 )   
References | Related Articles | Metrics
In this paper,considering kernel and parameter-less nearest-neighbor graph,a novel method named parameter-less supervised kernel locality preserving projection algorithm which aims at discovering an embedding that preserves nonlinear information was proposed for face representation and recognition.In this algorithm,firstly,by changing the Euclidean distance to the Cosine distance which is more robust to outliner,and constructing a parameter-less nearest-neighbor graph,this algorithm uses the nonlinear kernel mapping to map the face data into an implicit feature space.And then a linear transformation is preformed to preserve locality geometric structures of the face image,which solves the difficulty of parameter selection in computing affinity matrix.Experiments based on both ORL and Yale face database demonstrate the effectiveness of the new algorithm.
Method of Low Resolution Facial Fatigue Expression Recognition Based on Sparse Representation
ZHANG Ling, TIAN Xiao-lu, LUO Yuan, CHANG Jie and WU Yong
Computer Science. 2016, 43 (9): 305-309.  doi:10.11896/j.issn.1002-137X.2016.09.061
Abstract PDF(1562KB) ( 667 )   
References | Related Articles | Metrics
In order to effectively improve the performance of facial fatigue expression recognition on the low resolution image,a method of fatigue facial expression recognition based on sparse representation was proposed.Firstly,the reliability analysis method of Kendall coefficient of concordance is used to construct the low-resolution facial fatigue expression database TIREDFACE.Secondly,the sparse representation of the low resolution facial fatigue expression images of the identified test samples in the database is sought,and then the compressed sensing theory is used to seek their sparsest solution.Finally,according to the sparsest solution,the low-resolution facial fatigue expression classification is performed.Experimental results on TIREDFACE database show that the low resolution facial fatigue expression perfor-mance obtained by this method is much better than the linear classifier,the nearest neighbor (NN),support vector machine (SVM) and the nearest subspace (NS).Therefore,the proposed method on the low resolution facial fatigue expression recognition tasks achieves better performance and high accuracy.
Research on Background Model Adaptation for Acoustic Event Detection and Classification Based on Acoustic Surveillance System
ZHANG Ai-ying and NI Chong-jia
Computer Science. 2016, 43 (9): 310-314.  doi:10.11896/j.issn.1002-137X.2016.09.062
Abstract PDF(423KB) ( 615 )   
References | Related Articles | Metrics
Acoustic event detection and classification have become an important research problem as the increasing use of audio sensors in surveillance system.In these systems,audio circumstance is very complicated,that is,different locations,different noises,which cause the acoustic event detection and classification to be very difficult.Therefore,it is important to implement the background model adaptation in order to adapt these variations of background.In this paper,we proposed to use the constrained maximum likelihood linear regression (CMLLR) to adapt background model.Using the real world data and simulated data,we investigate the background model adaptation approaches and strategies for background model adaptation.Experimental results show that background model adaptation can improve the perfor-mance of target acoustic event detection and classification,and also can greatly reduce the false alarm of target acoustic event detection and classification.
Image Segmentation Method Based on Background Model and its Application in Face Recognition
WEI Lin-jing, NING Lu-lu, DAI Yong-qiang and HOU Zhen-xing
Computer Science. 2016, 43 (9): 315-319.  doi:10.11896/j.issn.1002-137X.2016.09.063
Abstract PDF(1531KB) ( 458 )   
References | Related Articles | Metrics
The foreground image is hard to be extracted when image’s intensity is inhomogeneous or the contrast is low.To solve this problem,an image segmentation method was proposed.This method reconstructs background model by combining sine basis functions with absolute distance measurement,and solves the model according to optimization theory and iteration method.This method discriminates between background and foreground of every pixel by comparing the intensity of each pixel in the background model with that in real image.To deal with the situation of inhomogeneous intensity of image,the image is divided into blocks before image segmentation,and is segmented according to background model and background similarity among adjacent blocks in sub-block image.Experimental results show that,comparing with classical methods including fuzzy C-means and OTSU,this method has lowest segmentation error,especially for the image with inhomogeneous intensity and low contrast.In the applications of palmprint image segmentation,comparing with iterated line tracing method and rough-fuzzy set method,this method has the characteristics of low error rate,high signal to noise ratio and shorter processing time.