Search

Aiming at the particle filter's inefficacy problem caused by narrow likelihood of accurate measurement system, a new particle filter with resampling step based on clustering and determination quasi-Monte Carlo is proposed. When importance samples are lost, the algorithm extracts key samples using clustering method, and then determines the number of spring-offs of them and sampling space according to the new support set composed of key samples. New particles are generated by QMC, in order to obtain more efficient samples and avoid the inefficacy phenomena. Simulations show that, compared to the general resampling step of particle filter this algorithm obtains more accurate and steady state estimation. [ABSTRACT FROM AUTHOR]

Sample selection can speed up the training of Fuzzy Support Vector Machine(SVM). However, it is difficult to select effective sample and the selection ratio is very high. This paper proposes a new sample selection method for Fuzzy SVM based on shadowed sets. We divide the fuzzy sets into three subsets, i.e. trustable data sets, trustless data sets and uncertain data sets. The samples are only selected in trustable data sets and uncertain data sets by using the subspace selection algorithm and the border vector extraction method respectively. Experimental results show that the training time and selection ratio is significantly reduced without any decrease in generalization ability by using the samples chosen by the proposed method. Furthermore, it improves the prediction performance of the classifiers when the data sets contain noises. [ABSTRACT FROM AUTHOR]

To deal with the problems such as high complexity and low accuracy of redundancy detection, a model of redundancy detection based on control structure analysis is proposed and implemented. This paper predigests the complexity of control structure by establishing a compound node table for tokens, which reduces the complexity of redundancy detection, and then detects the idempotent operations, dead code and redundant assignment. Experimental results of the open source code of Linux show that this model can find redundant code accurately and also has a low time-complexity. With this model, it is very convenient for developers to detect and correct these kinds of defects, and thereby to further guarantee the software quality. [ABSTRACT FROM AUTHOR]

To solve the problem of track association in dense target environment, a new track association algorithm based on K-Medoids clustering is presented. The algorithm makes the sensor tracks associate with the system tracks and selects the system tracks as Medoids, which greatly reduce the number of track pairs needed to associate, avoid the inherent defects of K-Medoids clustering, and improve the efficiency of the algorithm. In addition, the approximate distance between the two tracks is acquired by calculating the distance at a sampling point with the method of infinite norm, and then the track pair with minimum distance is selected as the associated tracks. The consideration of the historical and current track improves the probability of correct association. Finally, the algorithm is implemented in the multi-sensor and multi-target environment and the simulation results prove the effectiveness and superiority of the algorithm. The algorithm has a strong robustness in the presence of noise and outliers, and is suitable for the dense target environments. [ABSTRACT FROM AUTHOR]

There are many requirements for the multilanguage static information retrieval, and it is wasteful and complex to build specific front end for each language. So to meet the need and conquer the weakness we present a method based on GCC source code change to provide a uniform interface for retrieval. According to the static information type and the GCC inside mechanism, this method integrates the specific hook point with the GCC's internal functions to gather the needed. The details to be collected include: the type and function declaration, the statements traverse and the uniform multilanguage middle-representation. The reusability of this method reduces the duplicated cost of the construction for each language. The comparison experiments shows that this method can efficiently and robustly parse multilanguage and be directly applied to large -- scale open source code to retrieve the static information. [ABSTRACT FROM AUTHOR]

For solving the problem that the current clone and related bug detection tools cannot analyze largescale code and detect modified code clones, a model of clone and identifier inconsistency bugs detection based on the improved model of clone detection via sequential pattern mining is proposed. This model improves the existing submodel of identifier forgotten renaming bug detection, and introduces a new submodel of identifier mistaken renaming bug detection. Identifier renaming inconsistency bugs are detected by computing identifier unchanged ratio (UnchangedRatio). Experiment results indicate that this model can detect clone related bugs on the basis of code clone detection, reduce false-negative rate and false-positive rate, and is effective in detecting two kinds of identifier inconsistency bugs. [ABSTRACT FROM AUTHOR]

To evaluate the reliability of Multiagent system accurately, a synthetic evaluation approach considering communication reliability, unit reliability and system topology is proposed. Multiagent system structures are summarized; the influence of communication reliability and unit reliability to the reliability of centralized, distributed and federated Multiagent system is analyzed based on Markov process; and system reliability models are given. The simulation results show that this method can simplify the problem of complex system reliability, for the entire system reliability can be obtained after the reliability of every part is achieved. [ABSTRACT FROM AUTHOR]

In order to correctly analyze the control dependence and data dependence relations of C program, and then can out the program slicing, code redundancy and code duplication detection, a method to extract the static information of C program from GCC AST text was put forward. Firstly, the GCC AST text was standardized and the nodes unrelated to control dependence and data dependence analysis were eliminated. Secondly, control dependence subgraph was constructed. If data flow analysis was necessary, control flow graph was constructed based on control dependence subgraph and data dependence subgraph was constructed based on control flow graph. At last, by the introduction of inter-process analysis, system dependence graph was improved. It is indicated that the research can correctly analyze the control dependence and data dependence relations. [ABSTRACT FROM AUTHOR]

Aiming at the high computational complexity of particle filters, in order to reduce the number of samples, this paper proposes an improved particle filter based on subtractive clustering. Cluster vectors, composed of particles and their corresponding weights, are classified at a given radius through the improved sub-cluster algorithm presented by this paper, and then all the cluster vectors are replaced by the central vectors obtained from the classifying processing. Finally the central vectors are decomposed and the new particles mad their weights are restructured. The simulation results show that the proposed algorithm maintains the performance of the general particle filters, and meanwhile keeps less number of samples and higher computational efficiency. [ABSTRACT FROM AUTHOR]