Table of Content

15 October 2020, Volume 37 Issue 5 Previous Issue   

Select

Muti-scale Feature Fusion Network Based on Feature Pyramid Model GUO Qi-fan, LIU Lei, ZHANG Cheng, XU Wen-juan, JING Wen-feng 2020, 37 (5):  521-530.  doi: 10.3969/j.issn.1005-3085.2020.05.001 Abstract ( 204 )   PDF (3616KB) ( 351 )   Save Feature pyramid network (FPN) is an enhanced method for CNN network to express and output image information. It has been widely used in object detection network and has achieved significant effect improvement. The traditional feature pyramid model can not fully transfer the shallow details to the deep semantic features, which leads to inadequate feature fusion. It can only rely on the deep semantic information to make predictions, but ignore the underlying location information of the network. In terms of the above problems, we proposed a muti-scale feature fusion network based on feature pyramid model. Based on the FPN backbone, a mixed feature pyramid and a pyramid fusion module are designed. Based on the attention mechanism, multi-scale deep fusion of the feature pyramid is performed. We carry out the experiments on the PASCAL VOC2012 and MS COCO2014 datasets, and verify the effectiveness of MSFFN for feature fusion. Related Articles | Metrics

Select

Research on Census Omission Estimation HU Gui-hua, LIAO Jin-pen, FAN Shu-shan, YE Bao-hong, WU Ting 2020, 37 (5):  531-549.  doi: 10.3969/j.issn.1005-3085.2020.05.002 Abstract ( 147 )   PDF (272KB) ( 203 )   Save In view of many countries using unmatched estimator in census omission estimation, which leads to underestimate omissions, the objective of this work is to replace it with the census omission synthetic estimator. In order to achieve the goal, a combination of mathematical model and sampling estimation is used to study the unmatched estimator, census omission synthetic estimator, and their sampling variance estimators. Theoretical and empirical research results show that compared with the unmatched estimator, the census omission synthetic estimator provides more omissions and higher estimation accuracy. For simplicity of calculation, the census omission synthetic estimator must be established in population strata of equal probability. The census omission synthetic estimator will be applied to China's 2020 census omission estimation to improve its estimation accuracy. Related Articles | Metrics

Select

Optimal Reinsurance and Investment under $n$ Dependent Insurance Businesses YANG Peng, CHEN Xin 2020, 37 (5):  550-564.  doi: 10.3969/j.issn.1005-3085.2020.05.003 Abstract ( 117 )   PDF (249KB) ( 176 )   Save This paper investigates an optimal time-consistent reinsurance and investment problem with $n$ dependent insurance businesses for an insurance company. The insurance company is allowed to purchase reinsurance for reducing claim risk and invest in the financial market for increasing wealth. The financial market consists of one risk-free asset and $n$ correlated risky assets, whose price processes are described by diffusion processes. Then, we establish the wealth process of the insurance company by using the stochastic analysis theory. Our main goal is to find an optimal time-consistent reinsurance and investment strategy so as to maximize the expected terminal wealth while minimizing the variance of the terminal wealth. By applying the stochastic control and stochastic dynamic programming techniques, we establish the extended Hamilton-Jacob-Bellman (HJB) equation. Explicit solutions for the optimal reinsurance and investment strategy as well as the corresponding value function are obtained by solving the extended HJB equation. Finally, numerical experiments illustrate the effects of model parameters on the optimal time-consistent reinsurance and investment strategy. Related Articles | Metrics

Select

Analysis on Herding Effect of Agricultural Land Transaction Behavior LI Bo 2020, 37 (5):  565-582.  doi: 10.3969/j.issn.1005-3085.2020.05.004 Abstract ( 146 )   PDF (248KB) ( 224 )   Save It is the key link of establishing a unified agricultural land trading market, which is accurately evaluated the trading behavior of the transferors of agricultural land management right and effectively restrain the herding effect in agricultural land management right trading. Based on the individual rationality constraint of the agricultural land transfer, this paper establishes a sequential decision-marking mathematical statistical model, with the discrete behavior of two-dimensional uncertain signals as the spatial hypothesis. The process and mechanism of herd effect in the process of agricultural land transaction are discussed, and the generation of herd effect under different conditions is simulated. It is found that the herding effect of land transfer in agricultural land transaction is related to the possibility of obtaining information and signal. The main reason for herd behavior in agricultural land transaction is that the information is not transparent and accurate, which leads to a small probability of obtaining signals and makes it difficult for the transferors to judge the authenticity of the information. Related Articles | Metrics

Select

A Novel Linear Sampling Method for Reconstructing Scattering Obstacles from Data at One Frequency DING Meng 2020, 37 (5):  583-590.  doi: 10.3969/j.issn.1005-3085.2020.05.005 Abstract ( 95 )   PDF (175KB) ( 251 )   Save The theory and calculation of scattering and inverse scattering problems has always been an important subject in the field of applied mathematics, and its results have a wide range of applications in the fields of geological prospecting, non-destructive detection, medical imaging, etc. Linear sampling method (LSM) is a very popular non-iterative reconstruction algorithm in inverse scattering problems in recent years, but this method is difficult to generalize to more complex problems such as obstacle inverse scattering in half space.  This paper is associated with the numerical reconstruction algorithm in the inverse Dirichlet obstacle scattering problem from single frequency data. By constructing an auxiliary boundary value problem with an impedance boundary condition, a novel linear sampling method is proposed. Then it is theoretically proved that the method is effective for any given wavenumber in reconstructing the shape and location of the obstacle. Related Articles | Metrics

Select

The Construction of Object-oriented Multi-granularity Concept Lattice LI Ke-wen, LV Meng-meng, SHAO Ming-wen 2020, 37 (5):  591-605.  doi: 10.3969/j.issn.1005-3085.2020.05.006 Abstract ( 106 )   PDF (241KB) ( 200 )   Save How to reduce the complexity in the construction of concept lattice is an important research topic in formal concept analysis. Granular computing can analyze and solve problems from multiple angles which has been an effective tool for solving problems in the field of artificial intelligence. Aiming at the rapid construction of multi-granularity concept lattice, based on the granular computing theory, this paper investigates the relation between the extent and intent of concept before and after the change of the attribute granularity and defines different types of concepts. Then, Zoom algorithms composed of Zoom-in algorithm and Zoom-out algorithm are proposed, which achieve the transformation among object-oriented concept lattices with different attribute granularity combinations. Based on the original concept lattice, Zoom algorithms implement the construction of the new concept lattice directly, which avoid the trivial steps of the traditional construction method of concept lattice, and improve the efficiency of the construction of concept lattice. Therefore, the optimal granularity combination of the concept lattice can be quickly determined, which ultimately helps data mining and knowledge discovery from the data. Related Articles | Metrics

Select

An Adaptive Mesh Method Based on Radial Basis Function DUAN Xian-bao, DANG Yan, QIN Ling 2020, 37 (5):  606-614.  doi: 10.3969/j.issn.1005-3085.2020.05.007 Abstract ( 114 )   PDF (1070KB) ( 478 )   Save In this paper, we propose an adaptive mesh method based on the radial basis function. The proposed method uses the information of the numerical solution provided by the mesh dependence method and the difference solution from the radial basis function to refine or coarse the mesh. Our method takes full advantage of the simple format and the flexible node configuration that the radial basis function possessed and the robustness of the mesh-depended method. The proposed algorithm is easy to implement. The numerical examples show that the proposed algorithm can refine the mesh in the region where the solution changes dramatically and coarsen the mesh in the region where the solution changes gradually. Therefore, it can save an enormous amount of calculating time while ensuring the same numerical accuracy. Related Articles | Metrics

Select

A Filippov HIV/AIDS Epidemic Model Describing the Enhanced Diagnosis Measure Effect WANG Ai-li 2020, 37 (5):  615-630.  doi: 10.3969/j.issn.1005-3085.2020.05.008 Abstract ( 97 )   PDF (639KB) ( 385 )   Save A Filippov system has been proposed to describe the following diagnosis measure in HIV control: once the number of the diagnosed HIV infected individuals is above a certain level, the enhanced diagnosis measure is carried out; otherwise, a general diagnosis ratio is adopted. The sliding mode region and the sliding mode dynamics are explored by using equivalent control method. We also examine the global dynamics of the model based on the sliding mode dynamics and qualitative analysis. The main results show that the number of HIV infected cases can stabilize at a relatively high level or relatively low level if the threshold value is enough high or enough low. By choosing an appropriate threshold value, the number of HIV infected individuals fluctuates around this threshold level. Therefore, it is important to choose a threshold value properly for HIV control, according to which one can determine whether or not it is necessary to implement the enhanced diagnosis measure. Related Articles | Metrics

Select

Local RBF-FD Method for the Elliptic Eigenvalue Problem ZHANG Yi, FENG Xin-long 2020, 37 (5):  631-641.  doi: 10.3969/j.issn.1005-3085.2020.05.009 Abstract ( 125 )   PDF (334KB) ( 468 )   Save This paper is concerned with the application of the local RBF-FD method for the elliptic eigenvalue problem. The main idea of this method is to select the influential region corresponding to each interpolation node artificially and consider only the influence of the interpolation point in the region on the interpolation node, ignoring the influence of the nodes outside the region. This localization method obtains sparse matrices while losing a certain degree of computational accuracy, so that the algorithm can be applied to the computation of large-scale interpolation node scientific calculations. Through numerical experiments, we study the influence of node layouts, interpolation points, and the shape parameter on the calculation result of the eigenvalue problem, and using three radial basis functions to calculate and compare. The results obtained from the experiments are in good agreement with the analytical solutions to the problem. Related Articles | Metrics

Select

Stability and Stabilization of a Class of Fractional-order Nonlinear Delayed Systems ZHAO Dong, LI Ting-ting, YIN Hao 2020, 37 (5):  642-650.  doi: 10.3969/j.issn.1005-3085.2020.05.010 Abstract ( 164 )   PDF (164KB) ( 233 )   Save Stability is the foundation of analysis and design of the control systems. To explore the stability conditions of the fractional-order systems, especially fractional-order nonlinear delayed systems, is one of the most difficult issues in the field of the control theory and engineering. This paper is concerned with stability and stabilization of a class of fractional-order nonlinear delayed systems. By transforming the systems into an equivalent fractional integral ones and using inequality scaling technique, a new effective sufficient stability condition with simple form is established for such systems. Based on the proposed the results, some stabilization criteria are also derived by designing time-delay linear feedback controller. Finally, a numerical example illustrates the effectiveness of the results. In addition, the proposed methods are also extended and applied to other type fractional-order systems. Related Articles | Metrics