Computer Science

Review of Software Psychology

Computer Science. 2013, 40 (3): 1-7.

Abstract

PDF(747KB) ( 1314 )

RelatedCitation | Metrics

Since software is the production of human cognition,human is the key factor affecting software duality. With extention of "engineering system" to "socio-technical system"，software psychology increasingly manifests its implication value to software engineering. Firstly, the developing course and content framework of software psychology were summarized. Then the research advances in 6 branches were presented respectively,and the implications for software engineering were provided, and research shortages and clues for future study were analyzed. Finally, the overall developing trends of software psychology were concluded.

Survey on Embedded Urban Computing

Computer Science. 2013, 40 (3): 8-15.

Abstract

PDF(756KB) ( 666 )

RelatedCitation | Metrics

Internet of things realizes connecting of Internet with anythings by sensing technology. Cloud computing provides requirement oriented application service for users by sharing resources of flexible integration and dynamic configuration. The basic concept and function architecture was definited for embedded urban computing based on abovcmenboned background. Taking application scenarios of embedded urban computing as the studying object, this paper comparcel data collecting and service provision between Internet of hhings technology and traditional mode, disscrtated sensing application deployment, data capturing, information spreading for human interacting with environment, elaborated user-oriented data management service, sensing application service, location query service provided by cloud computing,analyzed privacy protection and identity authentication for RFID, key management for wireless sensor network, privacy anonymized for participatory sensing, trusted access control for cloud computing around the security and privacy problem for embedded urban computing,proposed the future research direction.

Survey on Fault-based Testing Techniques for Boolean Expressions

Computer Science. 2013, 40 (3): 16-23.

Abstract

PDF(753KB) ( 1021 )

RelatedCitation | Metrics

Boolean expressions play a key role in specifications and programs, and thus significantly affect the correctness of software. How to efficiently test Boolean expressions is an important issue. We first introduced the basic concepts and principle of fault based testing. We then summarized several fault classes of Boolcan expressions and discussed their detection hierarchies. Next, we reviewed several representative fault based testing strategics for Boolcan expressions,and proposed a comparison framework to compare them with respect to the applicability,fault detection capability and reduction of test suites. Finally, we pointed out several future issues based on our observations of limitations on the existing works,and briefly introduced our relevant work in this area.

Survey of Studies on Self-similarity in Storage System Wor(}load

Computer Science. 2013, 40 (3): 24-30.

Abstract

PDF(626KB) ( 826 )

RelatedCitation | Metrics

I/O bursty is one of the main reasons causing I/O bottleneck, so, it is significant for designing storage system and evaluating system performance to study and accurately synthesize the ubictuitous bursty in I/O workload. Research results show that the traditional poisson assumption is difficult to describe the I/+burstiness behavior well at the longterm time scales,and I/O bursty exhibits the similarity at different time scales,i. e.，self-similarity. So, self-similar models are used to characterize the long-range dependence in I/O workloads. Aimed at the Hurst parameter estimate, this paper summarized the time-domain and frectuency-domain estimators usually used to estimate the degree of self-similarity in storage workloads. After that, some existing models synthesizing I/O workloads were examined, thereinto, the characteristics of self-similar, multi-fractal and alpha-stable models were discussed. After summarizing the unresolved problems, this paper explored the future trend of the study on self-similarity in I/O workloads. The above work will provide a valuable reference for pushing the research on self-similarity in storage workloads.

Reconfigurable Instruction Set Processor Integrating Dynamic Sampling Profiler

Computer Science. 2013, 40 (3): 31-35.

Abstract

PDF(417KB) ( 700 )

RelatedCitation | Metrics

Reconfigurable instruction set processor (RISP) can dynamically extend its instruction set based on specific applications characteristics. The design of its hardware architecture and software tools is very different from the traditional design ideas. Based on the research for the features of RISP, a RISP integrating dynamic sampling profiler (DSP-RISP) was proposed. DSP-RISP has three kinds of working modes, obtaines the hot spots of applications through a hardware profiler with sampling mechanisms and then with the software tool chain semi-automatically completes the work such as instruction extension and generating, compiler retargeting and programmable hardware logic configuration to achieve the hardware adaptability and software compatibility for various embedded applications. Experimental results show that the sampling profiler of DSP-RISP is accurate and effective and under the limited hardware overhead, DSP-RISP can well adapt to application changes.

Design and Analysis of Protein Quantification Software Based on MPI and CUDA

Computer Science. 2013, 40 (3): 36-37.

Abstract

PDF(224KB) ( 778 )

RelatedCitation | Metrics

We introduced the design and the testing results of protein quantification software. The analytical results indicate that the testing size reaches 2048 cores, when P-QuantWiz runs on the clusters of DAWNING 6000，and the speed-up ratio still increases when we use 256 cores. Compared with the results of P-QuantWiz which is ran on one CPU, the speed-up ratio of PGQuantWiz on one single GPU is 8.1. Compared with the results which is ran on one GPU, the speed-up ratio of PGQuantWiz on 16 GPU is 14.18, and the parallel efficiency is 89%.

Load Balancing Strategy on Periodical MapReduce Job

Computer Science. 2013, 40 (3): 38-40.

Abstract

PDF(500KB) ( 745 )

RelatedCitation | Metrics

The MapReduce task load balancing in Hadoop mainly depends on the partition function. The Hadoop default partition function is not efficient in practical business processing. This paper presented a load balancing strategy based on the weight value of the periodic jobs. Because the data's distribution is similar in each period, we calculated the weight from historical data's profile. Through analyzing a sample data in Map phase to predict the whole data weighted integral approximate distribution, the strategy guids the Reduce partition to ensure its load balancing. We also presented the difference between TeraSort strategy and the new strategy. The experimental results with the view video logs show that the performance of our strategy is improved about 2 times compared with the default strategy.

Parallel Benchmark for Evaluating Parallel Simulation Engine

Computer Science. 2013, 40 (3): 41-45.

Abstract

PDF(405KB) ( 713 )

RelatedCitation | Metrics

SimK is a parallel discrete event simulation engine developed by state key laboratory of computer architecture in institute of computing technology, Chinese academy of sciences. Based on the released SimK-1.0, we extended the function of task partition and the blocking controlling in the process of synchronization. We released version 1.1 of SimK. On the other hand,as it lacks a benchmark to specifically SimK simulation performance and there is no comprehensive evaluation data,we first proposed the rules of development Benchmark for parallel simulation engine. Then we introduced the example "PassBa11". We used it to do the evaluation of SimK on the weak and strong scalability, as well as the strong scalability in unbalanced workload condition. Then we compared the speed-up ratio between the balanced workload and unbalanced workload condition in strong scalability test. The influence of simulated computing workload on speed-up ration from was also explored. We also discussed the applicability of the Benchmark. It can be concluded from our experiments as follows;a)our example "PassI3a11" is available to be the benchmark for SimK,as well as other parallel simulation engine. b) SimK has favorable strong and weak scalability. c) Both the load balance and the simulated computing workload will have effect on the speed-up ratio.

Efficient Sparse Matrix-vector Multiplication and CG Solver Optimization on GPU

Computer Science. 2013, 40 (3): 46-49.

Abstract

PDF(326KB) ( 1608 )

RelatedCitation | Metrics

Numerical methods of PDEs are mostly "compactly supported",say finite element,and finite difference methods etc. Due to the compact support, the global matrix associated with those numerical methods for scientific and engineering are sparse and very often also band shaped. We proposed and developed a high performance spMV algorithm for this specific but widely used sparse matrix type. The new algorithm, termed "bDIA (banded diagnal)", is implemented on NVIDIA GTX 285. Detailed comparisons with the five other mostly used sparse matrix formats/algorithms supported in the open source cuda linear algebra library (CUSP) show that bDIA doubles the best performance of the other five algorithms, breaking the float point efficiency limit of 4 0 o for single precision and 22.2% for double-precision. The conjugate gradient (CG) and the bi-conjugate gradient stabilized (BiCGStab) solvers both gain a speedup of around 1.5 using the proposed "bDIA" format/algorithm.

MapReduce-based Parallelization Model for Firewall Policy Conflict Detecting and Resolving

Computer Science. 2013, 40 (3): 50-54.

Abstract

PDF(423KB) ( 856 )

RelatedCitation | Metrics

Firewalls arc playing a very important role in network security, because the firewall policy rules arc determining that the network packet "Allow" or "Rejected" out of network. For the large networks,the rules are too many to ensure they have not conflict, therefore the detection and resolution of the policy conflict become an important aspect of network security. This paper presented a parallel method of firewall policy conflict detection and resolution algorithm,which resorts the segments formed by the rule-based segmentation technology, and translates the segments into the form of rules, uses this new rules instead of the original rules for packet filtering. Because all segments arc pairwise dis- jointed and every segment has one action, the conflicts in policy arc resolved.

Research on Non-intervention Information Acquisition and Public Sentiment Analysis System for Public Wi-Fi Wireless Networks Based on Supercomputer Platform

Computer Science. 2013, 40 (3): 55-58.

Abstract

PDF(367KB) ( 782 )

RelatedCitation | Metrics

An information acquisition and public sentiment analysis system for the city public Wi-Fi wireless networks was presented, which uses the petaflops computing platform in National Supercomputer Center in Ji nan. Based on the non-intervention wireless packets capture technology, Web page recovery and fault tolerant rcassembly technology, multiple text data mining technology and mass data process technology, the system can implement the functionality of wireless network forensics, public sentiment analysis and prediction, and provide overall and accurate references for the guidance of public sentiment for the government.

Parallel Multigrid Approach for Solving Poisson PDE in Gigapixel Image Editing

Computer Science. 2013, 40 (3): 59-61.

Abstract

PDF(304KB) ( 932 )

RelatedCitation | Metrics

With the development of image acctuisition technology, gigapixel images are being produced and emerged into the modern society, and how to efficiently compile these gigapixel images within gradient domain is the research focus of image processing and computer graphics. I}o solve the Poisson PDE with largcscale unknowns is crucial to gigapixel image editing in gradient domain. Traditional multigrid approach separately performs iteration, restriction and prolongation, bears heavy communication load between RAM and external memory. In the paper, a parallel multigrid approach for solving Poisson PDE was proposed, which exploits the locality and relevance of memory accessing and updating among the different stages to parallelly Perform the iteration, restriction and Prolongation in the sweeping window. Experiments of image stitching show that the presented method has the higher efficiency than the algorithms of successive over-relaxation, gauss-wider iteration and traditional multigrid.

.loin Algorithms Based on Fermi Architecture

Computer Science. 2013, 40 (3): 62-67.

Abstract

PDF(554KB) ( 753 )

RelatedCitation | Metrics

In column database, the join operation is still the most important and the most timcconsuming operation.UPUs' powerful computing capabilities can provide new optimizing solutions. This paper proposed a new Hash Join algorithm and a new Sort merge Join which are both based on Fermi architecture. These two new implementation approaches take full advantage of the new parallel cache hierarchy of Fermi, and successfully reduce the cache miss rate of the join operation. Combined with CUDA stream technology, the new join algorithms can effectively hide the data transfer delay between the main memory and global memory, when the join operation generates a large number of results. The experimental results show that Hash Join based on Fermi performs better when it deals with data skew, while Sort merge Join based on Fermi is more stable. Comparing with the implementations based on multi-core CPUs, these two algorithms are faster, speed up 2.4 times at most, and the new Hash Join even achieves 217M tuples per second when foreign key's dataset is high skew.

Fine-grained Parallel Multi-pattern Matching for Backbone Network NIDS

Computer Science. 2013, 40 (3): 68-73.

Abstract

PDF(614KB) ( 697 )

RelatedCitation | Metrics

As the network bandwidth continuously increases, the network security has been seriously threatened by malicious behaviors and risks. Network intrusion detection system (NIDS) is one of the efficient measures to cope with intrusion threats and protect information security, which employs pattern matching techniques to analyze incoming packs is and detect potential threats. However, pattern matching is such a compute-intensive task that most current techniques can't meet the demand of KIDS for backbone networks over lOGbps speed. We proposed a novel Bloom filter based approach for pattern matching, called PBPM (Parallel-Bloom-filter-based multi-Pattern Matching). PBPM employs multiple copies of the same Bloom filter to carry out parallel matching on different positions of the input text at the same time. The fine-grained parallel approach is able to skip multiple characters per clock when implemented on FPGAs, dramatically improving pattern matching performance. Experimental results on the rule set from Snort 2.9 show that the throughput of PBPM exceeds more than 20Gbps.

Fine-grained Parallel Multi-pattern Matching for Backbone Network NIDS

Computer Science. 2013, 40 (3): 74-76.

Abstract

PDF(281KB) ( 953 )

RelatedCitation | Metrics

Deep packet inspection is an advanced packet filter technology, widely used in the network firewall. Pattern matching algorithm based on regular expression has become one of the important methods in achieving DPI. Because the traditional regular expression matching algorithm has higher time complexity, it can not be used in real-time network security protection. In order to improve the real-time of regular expression matching algorithm, this paper proposed the parallel regular expression matches algorithm based on distributed storage. This algorithm counts step data, uses of the distributed storage, and realizes parallel processing. Experimental analysis shows that compared with the traditional serial algorithm, the proposed algorithm increases at least five times in matching speed, and can effectively reduce the time complexity of the algorithm, improve the matching efficiency.

On-demand File Transfer Algorithm Supporting Data-parallel over Multi-cluster

Computer Science. 2013, 40 (3): 77-78.

Abstract

PDF(258KB) ( 704 )

RelatedCitation | Metrics

Data-intensive applications often require the efficient transfer of large amounts of data over wide-area distribirted computing environments. In parallel processing, such data transmissions usually happen between data generation clusters and the data storage or processing clusters from time to time. This paper presented an on-demand file transfer algorithm,which aims at minimizing the makespan of overall file transfer rectuests. Taking advantage of the high transfer performance within a single cluster, the algorithm performs simultaneous transfer of different pieces of files to multiple nodes of a cluster when one node of the cluster requests the files. It employs multi-hop path splitting and multi patking methods in[1]to improve the transfer performance of a single source-destination pair. When several requests compete for the physical links, the algorithm evenly allocates the bandwidth resources upon the transmission load of each request, to make full use of the network bandwidth.

Research on Mean Shift Algorithm Using OpenCL on Multiple Many-core Platforms

Computer Science. 2013, 40 (3): 79-85.

Abstract

PDF(931KB) ( 1137 )

RelatedCitation | Metrics

As a general-purpose programming standard for multiple platforms, OpenCL has accelerated many applications. Due to the differences of different platforms in hardware and software environments, general optimization methods may not accelerate the application well for all. Taking the optimization of the mean shift algorithm on GPU and APU platforms as an example, the paper provided several insights on contributions of various optimization methods on different platforms. On one hand, we explored the architectures of different platforms. On the other hand, we compared the pros and cons of different optimization methods. Based on meticulous evaluations of the pros and cons, we looked for the optimal solution. Experimental results show that, on AMD 5850, Tesla C2050 and APU A6-3650 platforms, the optimized algorithm achieves 9.68 X，5.74 X and 1.27 X speedups, respectively, and 79.73 X，93.88 X and 2.22 X speedups comparcel to the serial version, respectively, and 1.27 X and 1.24 X speedups compared to the CUDA version OpenCV program for the first two platforms,respectively.

DDR:A Multidimensional Case Retrieval Optimization Algorithm

Computer Science. 2013, 40 (3): 86-88.

Abstract

PDF(318KB) ( 669 )

RelatedCitation | Metrics

In ontology-based case retrieval system,because the number of cases in the database can double increases with time, the efficiency of case retrieval continues lower. This paper presented a multi dimensional case retrieval algorithm-the DRR(Dimensionality Reduction of R-Tree) , the algorithm dimensionality reduces the case point of the multidimensional space to two-dimensional space point, uses a two-dimensional space points to represent the collection on behalf of the class case points, and biulds R-tree spatial index for this two-dimensional space point accelerates the retrieval efficiency and accuracy by the two search methods. It is proved that the method not only improves the accuracy of case retrieval, but also greatly improves the efficiency of case retrieval.

ilser-based FRC Evaluation Selection Model for Web Services

Computer Science. 2013, 40 (3): 89-94.

Abstract

PDF(518KB) ( 646 )

RelatedCitation | Metrics

Evaluation of non-functional properties (NFP) of Web services is an important premise to service selection.Current evaluation methods have not realized the evaluation of each NFP in accordance with its inherent character, nor have taken the complex requirements of users on multiple NFPs into consideration. Service which can best meet the demands of users will not be chosen based on existing Web services evaluation methods. To address this problem, a user-based FRC evaluation selection model for Web services was proposed. The proposed model considers all possible requirements on each NFP, and employs degree of membership, fuzzy judgment theory, probabilistic statistical method as well as the hOPSIS multi attribute decision making method based on the hosted mathematical property for each NFP.Fuzziness sub-model F and Randomness sub-model R are established, followed with Certainty sub-model C, which finally realizes comprehensive evaluation and selection of services. Simulation results demonstrate that the proposed model can effectively select an appropriate service according to the user's requirements.

Parallel Statistical Computing with R

Computer Science. 2013, 40 (3): 95-99.

Abstract

PDF(447KB) ( 992 )

RelatedCitation | Metrics

The ever-growing size and complexity of modern data sets in statistical analysis constantly challenge the capabilities of existing statistical computing methods. High-performance statistical computing is a promising strategy to address these challenges and begins to play an important role in finance, economy, and government and so on where statistical computing is dominant. A survey on state of the art in parallel statistical computing, especially with R and within the last four years was given. Furthermore, special attention on implementations of parallel statistical computing with R under different computer systems in a users’perspective was presented. Lest of synthetic code and real application show effectiveness of these implementations.

High Performance Massive Data Computing Framework Based on Hadoop Cluster

Computer Science. 2013, 40 (3): 100-103.

Abstract

PDF(334KB) ( 864 )

RelatedCitation | Metrics

HPC of massive data presents tremendous value. However, cloud systems still lack HPC computing power.This study improved the HPC ability of cloud computing technology by adding GPU to the cloud system. The proposed platform is based on Hadoop MapReduce programming model, and it defines some OpenMP like directives to annotate MapReduce program. The annotated code will try to be executed in parallel. A GPUClassloader was designed to convert annotated java code regions to CUDA code. With JNI,generated CUDA code and run on the GPUs. The computing resups of GPUs can be transferred back to the map function, in the end, the map function finishes the rest computing. The platform can support the user to complete CPU, GPU collaborative large-scale data parallel processing programming conveniently.

Research of Message Scalable Technology over InfiniBand Network

Computer Science. 2013, 40 (3): 104-106.

Abstract

PDF(622KB) ( 1650 )

RelatedCitation | Metrics

InfiniBand is one of the most promising network interconnecting technologies in HPC. Its reliable connection services support RDMA, atomic operations, etc, and are widely employed in MPI and other similar parallel programming models. However, the cost of message queues and buffers for reliable connections rises dramatically when the scale of parallelism increases. As a result, it becomes the bottleneck for InfiniBand in large scale applications. ho solve the problem, this paper provided shared receiving queue and extended reliable connection, then brought forward the group connection technology based parallel communication mode. On hand, the memory cost in the computing node is cut down for at least 100 times smaller. And the most amazing thing is that the cost of memory will be relatively constant when the scale of parallelism is upgraded.

Optimizing HPL Benchmark on Multi-GPU Clusters

Computer Science. 2013, 40 (3): 107-110.

Abstract

PDF(306KB) ( 1298 )

RelatedCitation | Metrics

The current high performance Linpack benchmark accelerated by GPU usually employs the performance-based dynamic load balancing algorism. However, this algorism does not perform well on multi-GPU cluster. The reason is that each GPU on this kind of node has a smaller calculating scale, and the gap of the total performance between GPU and CPU is larger. Therefore, this article proposed an experience-based dynamic load balancing algorism and a multi-GPU adaptive load balance algorithm. Besides, the article tested these two algorisms on multi-GPU cluster and it achieves a 6.3% acceleration compared with the latest NV)DIA's HPL accelerated by GPU.

Parallelism and Research on Functions with Continuously Independent Data and Intensive Memory Access Using OpenCL

Computer Science. 2013, 40 (3): 111-115.

Abstract

PDF(697KB) ( 843 )

RelatedCitation | Metrics

Continuously independent data type means when calculating the continuous elements of destination matrix,the used elements of source matrices are also continuous and there are no relationship among them. Intensive memory access function is the function that has less computation but a lot of data transfer operations. This paper took the bit wise function as the example, studied and implemented the parallel and the optimizing methods of the continuously independent data and intensive memory access function on GPU platforms. Based on the OpenCL framework, this paper studied and compared various optimizing methods, such as vectorizing, threads organizing, and instruction selecting, and finally used these methods to implement the cross-platform transfer of the bitwise function among different platforms.The study tested the function's execution time without data transfer both on AMD GPU and NVIDIA GPU platforms.On the AMD Radeon HD 5850 platform, the performance has reached 40 times faster than the CPU version in OpenCV library, 90 times faster on AMD Radeon HD 7970 platform, and 60 times faster on NVIDIA GPU hesla C2050 platform. On NVIDIA GPU `hesla C2050 platform,the speedup is 1. 5 comparing with the CUI}A version in C}enCV library.

Uniprocessor Performance Tuning of a Structured Grid Based Parallel CFD Application

Computer Science. 2013, 40 (3): 116-120.

Abstract

PDF(454KB) ( 706 )

RelatedCitation | Metrics

This paper optimized the performance of a high order structure grid based parallel CFI)(Computational Fluid Dynamics) application from a view of uniprocessor optimization. Performance critical variables were identified and trans- formed into constant parameters to enable compiler to apply specific high level optimizations. Multi-level data buffering was applied for the application's main data structures based on their structure and access characteristics, enabling the main computation codes to access these data more efficiently. Some loop transformations were applied to optimize the application's memory access performance. Performance evaluation was carried out on "TianhclA" parallel computer in- stalled at national super computer center in Changsha. Compared to the original code compiled by Intel compiler with the highest optimization levcl,the optimized code improves the serial performance for about 22. 2%一28. 9 0 0 for an 100 million grid points 2D acrofoil test case, and improves the parallel performance for about 13. 9%一20. 2 0 0 for an 112 million grid points delta aerofoil test case.

Architecture-aware Parallel Task Clustering Policy in Heterogeneous Computing

Computer Science. 2013, 40 (3): 121-125.

Abstract

PDF(420KB) ( 629 )

RelatedCitation | Metrics

Heterogeneous computing has been a trend of high-productivity computing. Matching between parallel task and architecture in heterogeneous computing becomes a key idea to realize high productivity. We provided parallel task clustering policy based on matching between parallel task and architecture. Firstly we gave the concept of high-produc- tivity and the problem of clustering on heterogeneous computing. Secondly after theoretically analyzing the relation be- tween heterogeneous matching and productivity, we gave the method of realizing respectively computing and structure matching. hhirdly we gave accordingly the architecturcaware parallel task clustering algorithm. Finally the simulation experimental results show that such algorithms can effectively realize heterogeneous matching and enhance the hetero- gcncous computing productivity.

Improved Parallel Orthodir(m) Algorithm

Computer Science. 2013, 40 (3): 126-127.

Abstract

PDF(227KB) ( 715 )

RelatedCitation | Metrics

By changing two inner product of vectors into several continuous inner products, this paper changed the data dependency and puted forward an improved Orthodir(m) algorithm(IOrthodir(m)). hwo algorithms share the same convergence. Theory analysis shows that when the processors reach a certain number, IOrthodir(二)acts better than Orthodir(二).Meanwhile IOrthodir(二)is better than Orthodir(二)in extensibility .Experimental results also show that IOrthodir(m) is superior to Orthodir(二).

CC$:A Parallel Programming Language for Distributed Many-core Platforms

Computer Science. 2013, 40 (3): 128-132.

Abstract

PDF(447KB) ( 868 )

RelatedCitation | Metrics

We presented a parallel programming language CC$.CC$is developed to reduce the programming com- plexity on distributed many-core systems. hhe programming model of CC$is based on Multi-13SP model, which ab- stracts distributed many-core systems into three levels. Data on CC$machines arc classified into 5 categories by data locality and shared regions, that is convenient for data access among different levels. CC$also presents sonic virtual di- rectives to describe data access logically. The programs on CC$machines are executed with Multi-I3SP super-steps. hhere arc four key features of CC$:unified programming style for all levels, built in multi-level common address spaces, description of data access requests with expressions, compilctime optimization for data transport. hhe experi- menu show that CC$is easy and effective.

Brief Exploration on Technical Development

Computer Science. 2013, 40 (3): 133-135.

Abstract

PDF(400KB) ( 786 )

RelatedCitation | Metrics

National supercomputing centers at China work use building thought of local government investigation, and market oriented application performing. Supercomputing resources are always applied at general applications, as the lo- cal govenment more focuses on the high-performance computing applications and services related to local business,rather than supercomputing working as strategical role in the traditional way. It is a long-term researching topic how to make the supercomputing powerful as a super-carrier active and applicable to benefit the technical innovation. Some challeng- ing technical issues suiting for the superomputing were discussed by taking domestic supercomputing center as the ex- ample, and some useful advises were addressed for applying international supercomputing center at local services.

Reliability-Cost Optimization Scheduling Model

Computer Science. 2013, 40 (3): 136-141.

Abstract

PDF(889KB) ( 668 )

RelatedCitation | Metrics

Due to the heterogeneity and dynamicity of grid resources,therc arc a lot of problems unsolved in grid task scheduling. To cope with the shortcoming of traditional reliability evaluation model, a new type of failure named dead- line-missing-failure was introduced to support reliability evaluation for tasks with deadline constraint. Considering the impact of local tasks, we used a stochastic service model to describe dynamic workloads of Grid resources. A corre- sponding calculation method and proof for resource reliability were proposed. Based on the proposed grid resource relia- bility evaluation modcl,a mufti objective grid task scheduling model was presented. hhe task scheduling was formulated as an optimization problem to maximize the reliability and minimize the cost. Fer the issue, a chemical reaction optimiza- tion algorithm, called CRO, was developed to solve the NP problem. Experimental results verify the performance of pro- posed reliability model and CRO algorithm is superior to GA and PSO for Grid reliability-cost optimization problem in quality and cffcicncy.

Research and Implementation of Column-based Database Schedule

Computer Science. 2013, 40 (3): 142-146.

Abstract

PDF(491KB) ( 813 )

RelatedCitation | Metrics

The heterogeneous computing platforms composed of mufti-core CPU and GPU have been a important devc lopment direction of current high performance computing. In order to improve the performance of column-base database query effectively and make full use of the computing resources of heterogeneous computing platforms,based on a set of defined columrrbase database primi-lives, this paper presented a set of primitive schedule methods, which include a primi- live execution mechanism, a CPU primitive schedule method based on dynamic programming and a GPU primitive schedule method based on GPU memory management mechanism. It maks that system can reasonably make use of mufti core CPU computing resources, effectively make use of locality of data in GPU memory and improve the overall performance. This paper tested several typical queries in TPGH benchmark programs, and the result represents that the CPU primitive schedule method lets the performance of queries more stable and GPU schedule method makes it faster. Meanwhile, through the experiments, we found that these column-base database schedule methods on heterogeneous computing platforms have some defects, which indicate the improvement direction for follow-up work.

Research and Realization on Visualization of Radar Detection in 3D Sea Battle-field

Computer Science. 2013, 40 (3): 147-150.

Abstract

PDF(609KB) ( 935 )

RelatedCitation | Metrics

Radar coverage visualization is an important part of battlefield situation display and necessary dependence of combat situation deduction and situation evaluation. For the visualization acquirements of 3D sea battlefield, the problem of radar detection under multi-interferences was researched, and then a visualization model of 3D Radar Coverage and corresponding render method were proposed. Combined with 3D sea battlefield environment,moving targets and inter- ferences of multi-jammers, dynamic rendering of radar coverage under multi interferences in 3D sea battlefield was im- plemented, which not only enriches the display context of battlo-field, but also enhances the power of command and deci- sion in 3D sea battlefield.

Cross-layer Routing Metric for Wireless Mesh Sensor Network

Computer Science. 2013, 40 (3): 151-154.

Abstract

PDF(433KB) ( 855 )

RelatedCitation | Metrics

For the reduirements of wireless mesh sensor network, a comprehensive measure of cross-layer routing proto- col HLS was proposed. HLS algorithm uses the link quality information at the bottom, distribution information of MAC layer and combines with hop counts to make comprehensive judgment for the alternative route to avoid lower quality link and heavy load nodes. In the pros(ss of route founding,link quality and node load thresholds arc set in the middle nodes and the delay forwarding operations are taken according to relevant information to reduce unnecessary routing o- verhcad. In the routing maintenance phase, passive and active methods arc used to detect the routing path comprchensivly. Simulation results of NS2 show that HLS has better performance in throughput capacity, end-to-end delay and network life time.

Mobile Sensor Network Node Fast Consensus Filter Localization

Computer Science. 2013, 40 (3): 155-158.

Abstract

PDF(324KB) ( 614 )

RelatedCitation | Metrics

Aimed at the dynamic properties of mobile sensor networks topology, a quickly kalman-consensus filter local- ization scheme was proposed. I}his algorithm only uses the RSSI value between the communication nodes to quickly ad- just fusion step on Metropolis rule. Under the situation of unknown networks topology, more precise localization is a- chieved steadily by kalman-consensus filter. From the simulation results, compared with Saber algorithm, the proposed scheme can depress the communication consummation and suits mobile sensor networks.

Priority-Queuing Model for Multiple Services Scheduling

Computer Science. 2013, 40 (3): 159-162.

Abstract

PDF(309KB) ( 1043 )

RelatedCitation | Metrics

The fast developing services in Internet pose challenges to the efficient transmission of various types of data, so an effective packet scheduling scheme is needed to meet the QoS constraints of heterogeneous services. In this article, the priority-queuing model was used to study the performances of various strategics based on delay sensitivity and par kets length. And the non-preemptive short packet first strategy was proved to result in the minimal overall delay. With different strategies for delay-sensitive and non-delay-sensitive services, an optimal priority-queuing model for the schedtr ling of multiple Internet services was designed based on above conclusions. The results from simulation experiments in NS-2 verified the superiority of the model. The results also demonstrated the features of packets queuing under different traffic scenarios and the law of variation for such performance indices as packet delivery ratio, throughput and average delay, which can be used to design effective measures for performance optimization.

Data Consistency Trust Evaluation of Wireless Sensor Networks Based on Multi-event Concurrent

Computer Science. 2013, 40 (3): 163-166.

Abstract

PDF(344KB) ( 592 )

RelatedCitation | Metrics

Currently, trust evaluation models mainly research the trust of nodes communication and few of these models research the trust of data consistency. However,trust evaluation of data consistency requires that sensing area happens only one event in a period of time,otherwise events can not be distinguished from each other accurately and legitimate node may be mistaken as malicious nodes. To resolve this problem, we proposed a method of trust evaluation of data consistency which is suitable for multi-event is concurrent. Simulation results show that, when multi-event is concurrent in sensing area, networks can distinguish events from each other accurately. It also can discover and exclude malicious nodes and reduce the false positive rate.

Dynamic Load Balancing Diffusion Algorithm with Neighbors Loading Awareness

Computer Science. 2013, 40 (3): 167-169.

Abstract

PDF(315KB) ( 601 )

RelatedCitation | Metrics

Dynamic load balancing is the primary means to adjust load distribution of network nodes. Compared with other load balancing algorithm, diffusion algorithm has advantage of performing synchronously, low migrating cost with- out center bottleneck, and so on. The paper proposed a dynamic loading balancing diffusion algorithm with neighbors loading awareness(NLA-LI3) to improve FOS diffusion algorithm. To realize convergence acceleration, sender initiated policy was adopted to establish a dynamic migrating group, and then information of neighbor loading was used to calcu- late loading exchanging impact factor. Experiment result shows that NLA-I3A's convergence speed can be improved ef- ficicntly with lesser migrating cost.

Network Coding Based Data Traffic Optimization in VANET

Computer Science. 2013, 40 (3): 170-174.

Abstract

PDF(796KB) ( 636 )

RelatedCitation | Metrics

The Vehicular Ad-Hoc Network is one of the important trends of the future network communications. hhrough VANE T, vehicles can not only share the traffic information to avoid accidents, but also exchange the multime- dia content through the peer to peer sharing technology. This paper is based on network coding technology to encode/ decode the file blocks in VANE?，clustering the vehicles according to their mobility, and combining the ROMSGP and modifying the priority method of the VANETC;ODE scheme in order to enhance the overall performance of VANETs. At last, we verified the effectiveness and correctness of this algorithm by simulation comparison of its performance with original algorithm. The simulation results show that the optimized VANETCODE has a significant improvement in the performance.

Bloom Filter Based Assessment Algorithm Supporting Service Fuzzy Matching

Computer Science. 2013, 40 (3): 175-179.

Abstract

PDF(429KB) ( 739 )

RelatedCitation | Metrics

The service search algorithms used in content based publish/subscribe system don't support service fuzzy matching. A service matching accuracy assessment method based on Bloom filter was presented. Facilitated by this ap- proach, an algorithm that supports service fuzzy matching was proposed. The main idea of this algorithm is using Bloom filter to describe the service and reduest, and assessing the similarity of service and reduest by the similarity of Bloom filter vectors. Experimental and theoretical results show that this algorithm can support content based service fuzzy matching by simple algebraic operations on Bloom filter. The evaluation accuracy rate is beyond 95 0 o.

Network Intrusion Detection Algorithm Based on Unsupervised Immune Hierarchical Optimization

Computer Science. 2013, 40 (3): 180-182.

Abstract

PDF(344KB) ( 648 )

RelatedCitation | Metrics

When the external network accesses university network, the external network data has no category tags, so the data recognition is unclear. I}he traditional intrusion detection model can not effectively extract the identifying char- acteristics of the unsupervised external network accessing data, and intrusion detection model can not be accurately trained, which makes accuracy of the college network intrusion detection is not high. I}o solve this problem, this paper proposed a intrusion detection algorithm based on unsupervised immune hierarchical optimization to learn data in the immune network,complete the data compression using the small-scale network,focus on improving the identifying cha racteristics of the data, and analyze the network using hierarchical clustering method to complete the establishment of the data model. Simulation results show that this unsupervised intrusion detection model method overcomes the obvious identifying characteristics of the university external network accessing data, and improves the accuracy of the university network intrusion detection,achieves satisfactory results.

Integrated Network Intelligent Node Deployment Algorithm of Improved

Computer Science. 2013, 40 (3): 183-186.

Abstract

PDF(322KB) ( 648 )

RelatedCitation | Metrics

In order to solve effects of intelligent node cross cover and mutual interference in complex environment inte- grated network system on intelligent node optimal position select, an integrated network intelligent node deployment al- gorithm of improved particle swarm optimization iteration was proposed. This algorithm first integrates network system intelligent node deployment model into the cross rate and interference constraint targets optimization problems,then u- ses the improved particle swarm optimization (pso) algorithm to map node deployment scheme, on the basis of the par- tide grain distance clustering degree and particle information entropy modifies particle weights, and then computes of the particle's fitness value, updates the local optimal solution and the global optimal solution of the particle,finally,u- ses the particle velocity and position update strategies to make iterative optimization to intelligent node deployment. Simulation experiments show that this algorithm has good convergence speed and convergence value is more excellent, effectively guarantees the maximum interference minimum coverage.

Analysis of Spreading Dynamics of Virus in Wireless Sensor Networks

Computer Science. 2013, 40 (3): 187-191.

Abstract

PDF(417KB) ( 774 )

RelatedCitation | Metrics

The spreading dynamics of wireless sensor networks ( WSN) was analyzed by considering the basic charac- teristics of WSN and the characteristics of sensor virus infection process. The different WSN models were established by employing different sensor deployment patterns, which include uniform random deployment, clustered deployment, group Gaussian deployment and random grid deployment. A virus spreading model was established based on infectious disease theory by considering the propagation characteristics of WSN and the impact of different MAC mechanism on virus spreading in WSN. The simulations show that the local group node deployment pattern would speed up the sprea- ding of virus in WSN, while the sleep/wake up communication mechanism would slow down the spreading of virus, re- duce the risk of the virus rapidly spreading in WSN.

SAMI. Path Verification Protocol for Web Service and its Performance Analysis

Computer Science. 2013, 40 (3): 192-196.

Abstract

PDF(389KB) ( 661 )

RelatedCitation | Metrics

To resolve the problem of increasing the length of SOAP message and reducing the respond speed of Web service seriously during the PKI-based signature to protect the SAML assertion, the paper proposed the SAML path verification protocol based on identity aggregate signature(II3ASPV),which improveds the respond speed of Web scrv- ice by shortening the length of the signature and the public key to reduce the transport time of SOAP message. I3y using SVO logic, we proved that the IBASPV protocol can ensure the integrity of the SAMI. assertion and source unforg(一 ability, and can protect the transmission path which can not be tampered with, and can prevent anti-replay attacks. I3y measuring the cryptographical calculation time and transport time based on Rampart module, we analyzed the trend that these factors which cause response time increase with the network data transport speed. Finally, we compared the per- formance of Web service based II3ASPV protocol with that based on PKI signature.

Formal Design and Verification of Interrupt Mechanism Based on Microkernel

Computer Science. 2013, 40 (3): 197-200.

Abstract

PDF(410KB) ( 822 )

RelatedCitation | Metrics

It is difficult to describe the correctness and security of the operate system (OS) by quantitative analysis. Formal method is the acknowledged standard one in design and verification for OS. Based on the operate system object semantics model (OSOSM) , we designed and verified the interruption mechanism of microkernel architecture using for- mal method,which was realized on our self-implemented verified trusted operate system (VTOS). Meanwhile,we used the theorem prover Isabelle/HOI. to formally describe the design process, and verify the integrality of the interruption mechanism of VTOS. Our research plays certain referential significance on formal design and verification of OS.

Improved Data-Cache Timing Attack on Cryptography Adopting Sliding

Computer Science. 2013, 40 (3): 201-205.

Abstract

PDF(429KB) ( 1263 )

RelatedCitation | Metrics

Publio-key cryptography, such as RSA and DSA, adopt sliding window method for modular exponentiation, from which sidcchannel information can be leaked while accessing Cache during execution, thus private key can be de- crypted. Exponent analysis algorithm is the key point to improve the efficiency of the attack. I3y analyzing the shortco- ming of previous work,this paper farther analyzed the relationship between Cacho-access trace and precomputed multi- pliers and proposed an improved exponent analysis algorithm based on window-value identifying. Experiments were made to prove the efficiency the improved algorithm and results showed that the improved algorithm was able to recover 60 0 o exponential bits, which is better than the previous result 47 0 o. In the end, the application of the improved algo- rithm was showed on RSA and DSA.

Model Based Automatic Fuzzing Script Generation

Computer Science. 2013, 40 (3): 206-209.

Abstract

PDF(302KB) ( 790 )

RelatedCitation | Metrics

Knowledge based fuzzing techniques generally have some shortcomings of heavy workload in writing scripts. A model based automatic fuzzing script generation method was proposed. Firstly data fom}at is represented by the higher order attribute grammars, and uniform data representation which is irrelevant to the test environments can be obtained. Secondly, the grammar model is used to parse the sample data and build the grammar parsing tree. Lastly, the relation- ship between the parsing tree and the test logic is built, which can be used to generate test script automatically. Experi- mental results indicate that the method can generate effective test scripts automatically to discover potential vulnerabili- tics in software.

BN Model Based Virtual Machine Security Deployment Policy in Cloud Computing

Computer Science. 2013, 40 (3): 210-214.

Abstract

PDF(676KB) ( 664 )

RelatedCitation | Metrics

CVDBN security deployment model of virtual machine was proposed in allusion to security risk caused by multiple consumers' sharing hardware resource in the environment of cloud computing and the security rules of subject server to deploy the object virtual machines were designed based on BN modcl,which can meet users' isolation require- ment on deployment of virtual machines with interest conflict. In the end, the security module of deploying virtual ma- chines and the security algorithm of deploying virtual machines were designed. Simulation was conducted which demon- strates the effectiveness and availability of the model and algorithm proposed in this context.

Design Model Synchronization Method for Software Product Line Based on GRoundTram

Computer Science. 2013, 40 (3): 215-218.

Abstract

PDF(433KB) ( 599 )

RelatedCitation | Metrics

The important products of the software product line domain-level and product level development periods domain design model and application design model, respectively. An automatic or semi automatic method is needed to keep synchronization between these two models during the life cycle of a software product line. Such a GroundTram- based method called SPI_Syn}GRound Tram was proposed in this paper. I}o keep synchronization between domain and application model, this method firstly converts this problem to a bidirectional transformation between two graphs, which is then solved gracefully by Ground Tram. The detailed process of this method was illustrated in this paper, and an on- lincshopping case model was used to show its effectiveness.

Attribute-Value-Distribution Based Result Ran(}ing Algorithm for Object-level

Computer Science. 2013, 40 (3): 219-224.

Abstract

PDF(564KB) ( 601 )

RelatedCitation | Metrics

Currently, there arc many ranking methods sorted by correlation in relational databases.()均cct level retrieval can integrate information which is dispersed in many tuples into complete information. As each object is unique, the difference among the objects is not only embodied in the keywords, but also embodied in the attribute values which they contain. Thus,this article described a method which gathers statistics of the distribution of attribute values in the object which contains the keyword, and makes use of the information entropy method to assign weights for each attribute value to calculate the correlation score of each tuple to each keyword. In the end, the object's score for each keyword is summed up to get the final ranking score for the object.

Method of Prioritizing Requirements Based on the Combination of ISM and AHP

Computer Science. 2013, 40 (3): 225-227.

Abstract

PDF(340KB) ( 1015 )

RelatedCitation | Metrics

In the methods of current requirements prioritization, the user requirements arc compared and prioritized on the same level, and the hierarchy of requirements is few considered. This paper proposed a method of prioritizing re- quirements based on the combination of ISM and AHP. First, the hierarchical model of software requirements was built with ISM. Next, the requirements of top level of the model were prioritize by analytical hierarchy process. Final, the case analysis was given. The results show that this method can improve the rationality of the prioritizing requirements, and reduce the times of comparing requirements,and reduce the failure probability of the project.

HBase Based Parallel BFS Method

Computer Science. 2013, 40 (3): 228-231.

Abstract

PDF(328KB) ( 1035 )

RelatedCitation | Metrics

As the next generation of storage model of giant data, NoSQL database plays an important role both in the fields of scientific computing and commercial computing, and has gained wide attention in academia and business community. We presented a new parallel method based on HBase to gain the shortest path tree. Firstly, Watts-strogatz mode was used to complete the mathematical modeling of giant network, therefore the network would have some cluster effects. Secondly,we made a simplification and improvement to the parallel breath-first search method,in order to improve its calculation efficiency. In addition, we designed and implemented a large number of experiments. According to the experiment results,we obtained the giant network shortest path trec,and verified the correctness and validity of the algorithm. Meanwhile,Contrast to the other path algorithm,we verified the efficiency of the algorithm.

Parallel Top-k Keyword Search Algorithm in Probabilistic XML　Documents

Computer Science. 2013, 40 (3): 232-237.

Abstract

PDF(478KB) ( 580 )

RelatedCitation | Metrics

Probabilistic XMI. can describe the uncertain data effectively, and Dewcy code is the most important encoding method for indexing probabilistic XML documents. But during the keyword search in big probabilistic XML documents,it takes much time for comparing the Dewey code of keyword index frectuently. To deal with the problem above, the probabilistic XML document was partitioned into several fragments, and a new Dewey encoding method of keyword index for probabilistic XML documents was designed, thus, a Parallel Top-k Keyword Search Algorithm (PTKS) was proposed. I}he experiment results show that the PTKS algorithm has low time complexity, especially, its efficiency is improved significantly when the structure of the document is complicated.

Cooperative-state Network and its Application in Software Modeling and Analysis

Computer Science. 2013, 40 (3): 238-243.

Abstract

PDF(499KB) ( 708 )

RelatedCitation | Metrics

Software modeling is a fundamental task of software activities and the core of model driven development process. The "Silver Bullet" type of software modelling is been pursued in software engineering. Based on the analysis of existing software modeling methods, the cooperative-state network was presented to achieve static and dynamic characterization of software system. First, the formal definition, graphical representation and interpretation of related concepts of cooperative-state network were given. Taking a Boiler control software system as example, the model construction algorithm and model reliability analysis were presented in detail. And also, some advantages of cooperativcstate network in software modeling and implementation were described,including easy understanding,one model describing the whole static and dynamic characteristics, executable, analysable, better visualization and consistency, etc.

Slide Window Based Ant Colony Algorithm for Online Grid Scheduling

Computer Science. 2013, 40 (3): 244-247.

Abstract

PDF(302KB) ( 714 )

RelatedCitation | Metrics

Task scheduling is a key problem of grid computing. Currently, a large number of researches focus on offline grid scheduling. Research on online scheduling is inadequate. Here a slide window based ant colony algorithm (SWbAC) for online grid scheduling was proposed. Two kinds of slide windows are used respectively to record recently received gridlet and the dynamic statuses of grid resources. These information are used to predict the future incoming gridlets and resources' statuses. hhe experiment platform is based on an extension for Gridsim, simulating the dynamic changes of grid resources' status. Compared with Online Min-Min algorithm, our algorithm performs 3%~10% better in view of average makcspan of gridlcts.

Adaptive Double-resampling Particle Filter Algorithm for Target Tracking

Computer Science. 2013, 40 (3): 248-250.

Abstract

PDF(319KB) ( 757 )

RelatedCitation | Metrics

Aiming at the problem that particle filter has particle degradation and depletion and the number of particle set is not adaptive to improve filtering accuracy and convergence rate, this paper proposed a adaptive double-resampling partide filter algorithm. This approach can solve the problems mentioned above. The algorithm first uses the observation innovation to determine the resampling of the particle distribution program, then conducts a rcsampling on the basis of the initial resampling .The second sparse resampling reduces the number of particles for updating using a particle mergrog technique.This can improve efficiency in the use of the particles, and avoid the use of excessive particles and increase the amount of calculation.The simulation results based on the DR/GPS show that compared with the traditional PF algorithm, the algorithm is effective to improve the filtering accuracy and stability.

Generalized Bacterial Foraging Optimization

Computer Science. 2013, 40 (3): 251-254.

Abstract

PDF(305KB) ( 709 )

RelatedCitation | Metrics

In order to improve the performance of Bacterial Foraging Optimization (I3F0) , a new swarm intelligence optimization algorithm, called the generalized bacterial foraging optimization (GBFO)was proposed, which has new chemotactic operation, is only composed of chemotaxis with group aggregation mechanism and adaptive strategy, and swarming is cancelled. The chemotactic loop with adaptive diffusion mechanism can improve ability of overcoming the "premature"，and healthiness is redefined to reduce the computational complexity. 10 complex Benchmark functions were tested. The simulation shows that the GI3F0 has better search ability and stability, solution quality and efficiency than other typical algorithm up to 80%~90%，70%~80% among test functions. The comparisons also show GBFO has excellent performance.

Research of Incremental Learning Algorithm Based on KKT Conditions and Hull Vectors

Computer Science. 2013, 40 (3): 255-258.

Abstract

PDF(330KB) ( 759 )

RelatedCitation | Metrics

Because the classical support vector machine is difficult to realize incremental learning flectly and rapidly when the number of training samples gets larger, this thesis proposed an incremental learning algorithm based on KKT conditions and hull vectors. This algorithm first selects the hull vectors which contain all support vectors. Next, it eliminates the useless samples among newly-added ones by using KKh conditions in order to reduce the number of training samples, then starts increment learning. The experimental results show that this algorithm not only guarantees the precision and good generalization ability of the learning machine, but also faster than the classical SVM algorithm.Therefore, it can be used in incremental learning.

Study of Weighted Association Rules Mining Algorithms Based on Timing and Interest Degrees Constraints

Computer Science. 2013, 40 (3): 259-262.

Abstract

PDF(342KB) ( 781 )

RelatedCitation | Metrics

In order to solve the problems that the frequent set information mining is not perfect in the association rule mining algorithm and the timing cycle influences mining of transaction set frequent item, a weighted association rules mining algorithms based on the timing and interest degrees constraints was proposed. This algorithm firstly uses sequcntial sliding function of the timing affairs set for probability estimation and weight assignment, then sets simplificalion according to the interests of the constraint function and pruning theorem transaction, makes a weighted frectuent transaction set extraction according to the degree of support and support expectations, lastly derives weighted associalion rules based on the confidence.The experimental results show that the algorithm can meet the users requirement in mining association rules quickly and effectively.

Image Reconstruction Method Using Smoothed-to and Dual-tree Wavelet Transform

Computer Science. 2013, 40 (3): 263-265.

Abstract

PDF(614KB) ( 692 )

RelatedCitation | Metrics

This paper presented a new algorithm for image reconstruction based on smoothed-to and dual-tree wavclet transform to save execution time and increase the quality of reconstruction. The dual-tree wavelet transform was used for image sparse and the Smoothed-Lo algorithm was selected for reconstructing the image. The experiment results show that DTSL0 improves the execution speed and the the quality and robustnesss of the reconstructed image.

Algorithm for SVM Ensemble Based on Rotation Forest and MultiBoost

Computer Science. 2013, 40 (3): 266-270.

Abstract

PDF(493KB) ( 881 )

RelatedCitation | Metrics

To improve the performance of ensemble learning, an ensemble algorithm which is a combination of Rotation Forest and Mu1tiI3oost was proposed as follows:To improve the diversity between classifiers,rotation transformation in rotation forest is introduced to model the new data set, and for higher accuracy, each classifier is trained by MultiBoost on the transformed data set. Finally,majority voting method is utilized to fusion the base classifiers' recognition results.To attest the validity, we made experiments on UCI data sets. The experimental results suggest that our algorithm can get higher classification accuracy.

Application of Improved Variable Precision Rough Set in Concept Lattice Construction

Computer Science. 2013, 40 (3): 271-274.

Abstract

PDF(693KB) ( 635 )

RelatedCitation | Metrics

The algorithm of concept lattice reduction based on variable precision rough set was proposed by combining the algorithms of汗upper and lower distribution reduction in variable precision rough set with the characteristics of the formal context in concept lattice. The traditional algorithms about value select algorithm, attribute reduction based on discernibility matrix in VPRS were discussed. There are defects in these traditional algorithms which are improved. Finally, the generation system of concept lattice based on variable precision rough set was designed to verify the validity of the improved algorithm. And a case demonstrates the whole process of concept lattice construction. The experimental results indicate that this algorithm has great validity.

Stability Analysis of Particle Swarm Optimization Algorithm and its Improved Algorithm

Computer Science. 2013, 40 (3): 275-278.

Abstract

PDF(299KB) ( 1641 )

RelatedCitation | Metrics

The loss of population diversity is an important reason which leads to the premature convergence of the PSO algorithm Therefore, the stability of the PSO algorithm based on the linear time-invariant discrete system was analyzed theoretically and the possible reasons of the lack of the population diversity were discussed in this paper. Based on the stability of the algorithm, an improved PSO algorithm was presented in which the inertia weight factor value is got randourly within a certain range and the leaning factor value is a constant. In the algorithm the population diversity can be maintained by the character that the particle speed has certain probability. The simulation experiments of three constraint optimization problems show that the algorithm has great probability to jump out of local extremum,and avoids the precocious premature convergence effectively.

Hybrid Artificial Bee Colony Algorithm for Solving High-dimensional Function Optimization Problems

Computer Science. 2013, 40 (3): 279-282.

Abstract

PDF(332KB) ( 693 )

RelatedCitation | Metrics

In order to enhance the global search ability of artificial bee algorithm in solving complex function optimizalion problem,a hybrid artificial bee colony algorithm (HABC) was proposed. HAI3C is based on multi-parent crossover　and differential evolution, and the key points of it lie in;1) employs multi parent crossover to enhance the global search　capability of the algorithm;2) removes identical individuals from the population for maintaining the diversity;3) adopts differential evolution operator to speed up the evolution. Experimental results on high-dimensional function optimization problems show that HABC possesses more powerful global search capability and better convergence rate.

Real-time Anomaly Detection Based on Sampled Peculiarity Factor

Computer Science. 2013, 40 (3): 283-286.

Abstract

PDF(329KB) ( 710 )

RelatedCitation | Metrics

Peculiarity factor is an important concept in the peculiarity-oriented mining, but its computation is too complex. Using sampling-based method to define the peculiarity factors called sampled peculiarity factor (SPF) can improve operational efficiency without affecting the accuracy. To improve the efficiency of anomaly detection algorithm based on the SPF, learning sampling frequency algorithm was proposed. Combined the optimal sampling frequency and SPF, rcaltime anomaly detection algorithm was proposed. Experiments use real data sets, take confidence as 95 0 o，and the optimal sampling frequency sequence is between 1/32 and 1/16. Simulation results show that false detection rate of the algorithm is 2%.

Research on Differential Privacy Preserving k-means Clustering

Computer Science. 2013, 40 (3): 287-290.

Abstract

PDF(325KB) ( 1559 )

RelatedCitation | Metrics

We studied k-means privacy preserving clustering method within the framework of differential privacy. We first introduced the research status of privacy preserve data mining and privacy preserve clustering, briefly presenting the basic principle and method of differential privacy. To improve the poor clustering availability of differential privacy k-means, we presented a new method of IDP k-means clustering and proved it satisfies E-differential privacy. Our experimenu show that at the same level of privacy preserve, IDP k-means clustering gets a much higher clustering availability than differential privacy k-means clustering method.

Marginal Neighborhood Nullspace Discriminant Analysis

Computer Science. 2013, 40 (3): 291-294.

Abstract

PDF(322KB) ( 604 )

RelatedCitation | Metrics

A marginal neighborhood nullspace discriminant analysis was proposed. The proposed method firstly defines the objective function,and then gives the theory analysis and proof of the objective function. Therefore,this paper pointed out that the algorithm must firstly projects high-dimensional samples into low-dimensional subspace by using PCA algorithm as the first step. In the low-dimensional subspace, the objective function does not lose any effective discriminant information. hhis algorithm can effectively not only resolve the small sample size problem but also work out the orthogonality projection matrix only by the three eigenvalue decomposition. Finally, the nonlinear marginal neighborhood nullspace discriminant analysis based on kernel mapping was given. Experimental results on face database demonstrate the effectiveness of the proposed method.

New 2D Otsu Image Segmentation Method via Particle Swarm Algorithm and Adaptive Weighted Window

Computer Science. 2013, 40 (3): 295-298.

Abstract

PDF(581KB) ( 633 )

RelatedCitation | Metrics

Aimed at the shortage of the abilities of noise removing and small target preservation for the conventional two-dimensional Otsu thresholding method, a new two-dimensional (2D) Otsu method based on adaptive weighted window was proposed. hhe new method improves the window setting method of the 2D Otsu, and the window size is adaplively determined by the local stationarity character. Then, the threshold is computed by the particle swarm algorithm, in order to improve the segmentation performance and shorten the computational time. Compared with the commonly-used oncdimensional Otsu, 2D Otsu method and linctype threshold 2D Otsu method, the proposed method has the better segmentation performance, with better performance for noise removal and small target preservation.

Enhance Algorithm for Fog Images Based on Improved Multi-scale Retinex

Computer Science. 2013, 40 (3): 299-301.

Abstract

PDF(858KB) ( 977 )

RelatedCitation | Metrics

When images are shot in fog circumstances, the color and contrast of fog images appear degrade phenomenon.In order to improve the quality of fog image, this paper proposed a fog image enhancement method based on improved multi-scale Retinex algorithm. Firstly, the dynamic range of image was compressed by the power transform, and then nonlinear transform was used to suppress the high light area of image,finally the unsharp mask filtering was used to eliminate fuzzy to enhance the detail information of fog image,and a few fog images were used to test the performance of the proposed algorithms. The simulation results show that the proposed algorithm solves the shortcomings of traditional Retinex algorithms, accelerates the speed of fog image enhancement, and the fog image is more clearly, so it can obtain better visual effect

Hybrid SUSAN Algorithm and Robert Algorithm for Image Edge Detection Filtering Technique

Computer Science. 2013, 40 (3): 302-304.

Abstract

PDF(512KB) ( 712 )

RelatedCitation | Metrics

This paper studied the algorithms of image edge detection and filtering processing. Because the traditional SUSAN algorithm for threshold selection is difficult to get more information of the image, this paper put forward a Robert based on image edge detection technology improved algorithm, which combines with SUSAN feature point matcking technology,at the same time,uses average filtering algorithm to remove the noise in image detection process, and finally uses the image thinning method to do image refining processing. The simulation results show that the proposed algorithm can effectively detect the image and reduce the complexity of the algorithm.

Association Algorithm of Group Target in Low-resolution Remote Sensing Images Based on Relative Shape Context

Computer Science. 2013, 40 (3): 305-309.

Abstract

PDF(708KB) ( 668 )

RelatedCitation | Metrics

One of the important precondition of identification fusion based on remote sensing images is target association,which is to determine if the information from two or more images are related to the same target. A novel and robust point pattern matching method was presented for group target association in low-resolution remote sensing images.A new point set based invariant feature, Relative Shape Context (RSC) , was proposed. We used the test statistic of relafive shape context descriptor's matching scores as the foundation of mathematics model of group target association. For resolving the modcl,we firstly constructed the new compatibility measurement and used it to initialize the association probability matrix. Then the association probability matrix can be updated by relaxation labeling. The oncto-one matcking can be achieved by dual-normalization of rows and columns in the end. Experiments on both synthetic point sets and on real world data show that the group association algorithm is effective and robust.

Method of Medical Image Fusion Based on Nonsubsampled Contourlet Transform

Computer Science. 2013, 40 (3): 310-313.

Abstract

PDF(864KB) ( 685 )

RelatedCitation | Metrics

This paper proposed a novel method of medical image fusion based on nonsubsampled contourlet transform (NSCT) against the existing problems of medical image fusion by traditional multi-scale transform. Considering regional relativity of the adjacent low frectuency sub-band,a fusion rule based on local area energy was adopted in low frequency sulrband coefficient according to characteristics of medical image. When choosing the bandpass directional sulrband coefficients, the paper made best use of directional characteristics of NSCT. A fusion rule based on sum-modified-laplacian (SMI)was presented in bandpass directional sulrband coefficients. The experiment results show that the proposed method can avoid image distortion and achieve a good effect of image fusion compared with traditional fusion methods.