Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2022-05-18 Cooperative journals: 《计算机应用研究》
Abstract： It is difficult to predict the correct human poses when facing the challenge of the scale change of the feature map in the human pose estimation. To solve this problem, proposing a high-resolution network MSANet (Multiscale-Attention Net) based on multi-scale attention mechanism to improve the detection accuracy of human pose estimation. Introduce lightweight pyramid convolution and attention feature fusion to achieve more efficient extraction of multi-scale information; citing the self-transformer module in the fusion of parallel subnets for feature enhancement to obtain global features; in the output stage, The features of each layer are fused using an adaptive spatial feature fusion strategy as the final output, which more fully obtains the semantic information of high-level features and the fine-grained features of low-level features to infer invisible points and occluded key points. Tested on the public dataset COCO2017, the experimental results show that this method improves the estimation accuracy by 4.2% compared with the basic network HRNet.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2019-04-01 Cooperative journals: 《计算机应用研究》
Abstract： Aiming at the discontinuity of the traditional hard threshold function at the threshold and the constant deviation between the original wavelet coefficient and the wavelet estimation coefficient in the soft threshold function, an image denoising algorithm based on the improved threshold function is proposed. The algorithm combines the advantages of the improved threshold function, dynamically selects the fixed threshold by setting appropriate adjustment parameters, and adds the adjustment factors to reduce the constant deviation between the original wavelet coefficient and the estimated wavelet coefficient, thereby improving the degree of approximation of the reconstructed image and the original image. The improved threshold function satisfies continuity at the threshold while satisfying the asymptotic and higher order conductibility of the function. Simulation results show that the improved threshold function is used for image denoising. There is a good visual effect. On the other hand, the denoising performance indicators such as mean square error (MSE) , peak signal-to-noise ratio (PSNR) and signal-to-noise ratio (SNR) are performed. In comparison, both PSNR and SNR are improved, MSE is reduced. The denoising effect is optimized.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2019-01-28 Cooperative journals: 《计算机应用研究》
Abstract： In order to improve the performance of whale optimization algorithm for solving complex function optimization problems, this paper proposed an improved whale optimization algorithm based on adaptive parameters and niche technology. Firstly, the algorithm introduced an adaptive probability threshold to coordinate the exploration and exploitation ability. Then, the algorithm used adaptive position weights to adjust the whale position update formula to improve the convergence speed and search precision. Finally, the algorithm used preselection niche technology to avoid premature convergence. The results on 12 typical benchmark functions shows that the improved algorithm has faster convergence speed and higher search precision than other comparison algorithms. It proves that the improvement strategy can effectively improve the performance of the whale optimization algorithm for solving complex function optimization problems.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2019-01-28 Cooperative journals: 《计算机应用研究》
Abstract： However, there are still some shortcoming in cuckoo search algorithm(CS) , such as low convergence precision, slow convergence speed, Weak search vitality and interference phenomena among dimensions when dealing with high-dimensional optimization problems. Dynamically Adaptive Cuckoo Search Algorithm Based on Dimension by Opposition-based Learning(DA-DOCS) was proposed, Firstly, the selected solution updated for dimension-by-dimension by Opposition-based Learning, this result reduced interdimensional interference and expanded population diversity. Then the method of elite retention was used to evaluate the results and improve the search ability of the algorithm. Finally, the information of the current solution was fully utilized to dynamically adaptive the scaling factor control to guide the solution to converge quickly and enhance the search vitality of the algorithm. The experimental results show that compared with the standard cuckoo search algorithm, the proposed algorithm has improved convergence precision, convergence speed and search vitality. Compared with other improved algorithms, it has certain competitive advantage.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-12-13 Cooperative journals: 《计算机应用研究》
Abstract： According to the slow convergence speed and low searching precision of traditional backtracking search optimization algorithm, this paper proposed an improved algorithm based on cellular automaton and orthogonal experimental design. Firstly, the algorithm introduced orthogonal experimental design method into the crossover operator to obtain representative high-quality offspring individuals; Then based on the neighbor model of cellular automaton, the method carried out the orthogonal crossover operation of multiple parents in the domain for individuals, which was beneficial to improve the mining capacity and search efficiency of the algorithm; Finally, in order to balance the global searching and local searching ability of the algorithm, the method introduced the dynamic proportional weight of excellent individuals into the cross-population to select and update them, with adopted a new dynamic variation equation. The simulation experiments selected 12 standard test functions and compared with 6 other well-behaved algorithms, the results show that the improved algorithm has obvious advantages in convergence speed and optimization accuracy.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-10-11 Cooperative journals: 《计算机应用研究》
Abstract： The accuracy of traffic flow forecasting plays an important role in the field of Intelligent Transportation Systems. In order to improve the accuracy of traffic flow forecasting model based on Least Squares Support Vector Machine, this paper proposed a novel modified gravitational search algorithm (TCK-AGSA) for parameters optimization. Firstly, this paper improved the Kbest function based on Tent map, so that the algorithm has a mechanism to jump out of local optimum. Then, by introducing the guidance of global optimal to accelerate the movement of agents towards optimal solution. Furthermore, it introduced the evolutionary factor and converge factor into the weighted coefficient of agent’s velocity to make the algorithm more adaptive. The simulation results for 12 benchmark functions show that the performance of TCK-AGSA is better than GSA and its variants. Finally, this paper proposed a LSSVM model optimized by TCK-AGSA, and selected the 2016 actual traffic flow data of Guizhou Expressway for experiment. The results show that the proposed model has better prediction accuracy, robustness, and generalization capability.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-10-11 Cooperative journals: 《计算机应用研究》
Abstract： In order to solve the disadvantage of the traditional whale optimization algorithm, which is slow convergence and easy to fall into local optimum, this paper proposed a mixed strategy based whale optimization algorithm. Firstly, it introduced the nonlinear adjustment strategy to modify the convergence factor, balance the exploration and exploitation capability and accelerate the convergence speed. Then, by introducing an adaptive weighted coefficient into the position update formula of whales to improve the search precision of the algorithm. Finally, it combined the limit threshold idea of artificial bee colony algorithm to effectively jump out of the local optimum and prevent premature convergence. The results show that the proposed algorithm has better search precision and convergence speed through experiments on different dimensions of 14 benchmark functions.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-08-13 Cooperative journals: 《计算机应用研究》
Abstract： In order to solve the problem that the clustering result of K-means algorithm gets affected by the initial cluster centers easily, this paper proposed a novel K-means clustering algorithm based on improved gravitational search algorithm. Firstly, it enhanced the global exploration and local exploitation capability of the algorithm with the introduction of adaptive concept to control the attenuation factor of gravitational constant. Then, by introducing immune clonal selection algorithm to make the algorithm jump out of the local optimum efficiently. The experimental results on twelve test functions prove the effectiveness and superiority of the improved GSA. Finally, by combining the improved GSA with K-means algorithm, this paper proposed a new clustering algorithm called A2F-GSA-Kmeans. The experimental results on six test datasets show that the algorithm has better clustering quality.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-07-09 Cooperative journals: 《计算机应用研究》
Abstract： According to the scarcity of wireless spectrum resources and low utilization rate, this paper presented a spectrum allocation method based on binary fireworks optimization algorithm. Each firework individual carries out distributed explosion search, and the method proposed an improved formula to update the explosion radius of the optimal fireworks dynamically. For the lack of information exchange between particles in the variation process, the algorithm introduced the crossover operator and mutation operator of the genetic algorithm to further enhance the population diversity. In order to avoid falling into a local optimum, the algorithm used simulated annealing with Metropolis criterion to perturb the selected optimal individuals. The simulation experiment shows that the method has the characteristics of high optimization accuracy and high convergence speed. It can achieve the maximization of the network efficiency and the proportional fairness among different users.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-05-20 Cooperative journals: 《计算机应用研究》
Abstract： Efficient access to the inverted index is a key aspect for a search engine to achieve fast response times to users’ queries. While compression of its posting lists is one of the most important methods to improve the performance of search engine. Segmentation method optimized by ABC algorithm in ASCS algorithm was proposed for the problem of ASCS algorithm that it adopts uniform segmentation instead of optimal segmentation; The ASCS algorithm only considers an influencing factor and ignores the influence of other factors ; The ratio of compressing long sequence of uneven distribution is unsatisfactory with ASCS algorithm, it was adopted that process integer sequence with sorting and differential encoding before ASCS algorithm. Simulation experiments show that the improved algorithm has significantly increased compression ratio of inverted index file comparing with ASCS algorithm.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-19 Cooperative journals: 《计算机应用研究》
Abstract： As a new kind of heuristic swarm intelligence algorithm, the Bird Swarm Algorithm (BSA) easily falls into problems about local optimal, slow convergence speed and low resolution accuracy. Considering the fact that the original Bird Swarm Algorithm is not sufficient to solve the issue in terms of optimization, this paper proposes an optimization algorithm, Dynamic Inertia Weight-Bird Swarm Algorithm (DBSA) . The algorithm corrects birds flying interval by introducing nonlinear dynamic inertia weight, balancing the abilities of population global search and local search; this paper introduces the parameter of levy flight in the process of simulation of the foraging birds producer, advancing algorithm’s vitality and effectiveness via replacing the original algorithm producers foraging strategies. As a result, experiments show that the modified algorithm improves the convergence speed and optimization accuracy effectively.
Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2018-04-17 Cooperative journals: 《计算机应用研究》
Abstract： With the rapid increase in the big data environment, frequent itemsets data mining faces in the actual incremental update problem. This paper proposes a parallel incremental updating algorithm based on MapReduce for frequent itemsets in frequent items ultrametric trees. The algorithm utilizes the support of frequent check ultrametric tree leaf node to determine the frequent itemsets and frequent itemsets using quasi strategies to optimize the parallel computing process, so as to improve the efficiency of data mining. According to the compared experiment results, it shows that the proposed algorithm is able to scan and update data efficiently, and has good scalability. It can be used for mining association rules in the incremental big data environment.
Hits
Hits
Downloads
Comment