BP神经网络权值初始值与收敛性问题研究

BP神经网络的收敛性问题是一直受到广泛关注的问题。本文针对BP网络在运算过程中陷入局部最小区，收敛速度慢的问题，从BP算法的原理出发，讨论了权值初始值对网络训练速度的强烈影响（仿真结果证明了这一点），并提出了解决方法。     

一种新的基于神经网络的高精度电力系统谐波分析算法

提出了一种新的基于三角基函数的人工神经网络算法,利用该算法可一次性获得电力系统基波及各次谐波的频率、幅值和相位,提出并证明了该神经网络算法的收敛定理,给出了利用该算法进行谐波分析的仿真实例.仿真结果表明,文中提出的谐波测量方法的计算精度极高,且计算量较小,因此在电力系统谐波测量中有较高的应用价值.     

3D ECT reconstruction by an improved Landweber iteration algorithm

Electrical capacitance tomography (ECT) is a relatively mature non-invasive imaging technique that attempts to map dielectric permittivity of materials. Recently, 3D ECT has gained interest because of its potential to generate volumetric images. The study of a fast and accurate image reconstruction algorithm is a challenge task, especially for 3D reconstruction. In this paper, we propose an improved Landweber iteration algorithm. We incorporate an additional acceleration term into the cost function and apply an adaptive threshold operation to the image obtained in each iteration for reducing artefacts. The algorithm proposed is tested by the noise-free and noise-contaminated capacitance data. Sensitivity matrixes and capacitance data of a 3D ECT sensor are obtained by using the finite element (FE) method. Extensive simulations in 3D reconstruction are carried out. The results verify the effectiveness of these improvements. Both the reconstruction time and the artefacts in the reconstructed image are reduced obviously. The experimental results of 3D reconstruction of objects in the shape of letters U and L confirm the effectiveness of the proposed algorithm further.     

Two-stage iterations based on composite splittings for rectangular linear systems

In this paper, we introduce a two-stage method to solve rectangular linear systems that exhibits faster convergence than typical stationary iterative methods. Under suitable conditions, we prove convergence of the new method. The number of outer iterations can be reduced by using a few significant number of inner iterations for efficient computations. Further, we perform a comparison analysis, and establish that a higher number of inner iterations ensures a smaller spectral radius of the global iteration matrix. We also discuss the uniqueness of a proper splitting, and illustrate different comparison theorems for different subclasses of proper splittings.     

Low climate stabilisation under diverse growth and convergence scenarios

In the last decade a number of papers have analysed the consequences of achieving the greenhouse gas concentration levels necessary to maintain global temperature increases below 2 °C above preindustrial levels. Most models and scenarios assume that future trends in global GDP will be similar to the growth experienced in the past century, which would imply multiplying current output by about 19 times in the 21st century. However, natural resource and environmental constraints suggest that future global economic growth may not be so high. Furthermore, the environmental implications of such growth depend on how it is distributed across countries. This paper studies the implications on GHG abatement policies of low global GDP growth and high convergence levels in GDP per capita across countries. A partial equilibrium model (POLES) of the world's energy system is used to provide detailed projections up to 2050 for the different regions of the world. The results suggest that while low stabilisation could be technically feasible and economically viable for the world in all the scenarios considered, it is more likely to occur with more modest global growth. However, that will imply higher global abatement costs relative to GDP. Convergence in living standards on the other hand places greater pressures in terms of the required reduction in emissions. In general we find that there are major differences between regions in terms of the size and the timing of abatement costs and economic impact.     

MAC finite difference scheme for Stokes equations with damping on non-uniform grids

In this paper, we consider the numerical methods for stationary Stokes equations with damping. The mark and cell(MAC) method has been applied to discretize the problem on non-uniform grids. We establish the LBB condition and the stability for the MAC scheme. Then we obtain the second order super-convergence in L2 norm for both velocity and pressure on non-uniform grids. We also obtain the second order convergence for some terms of H1 norm of the velocity, and the other terms of H1 norm are second order convergence on uniform grids. Numerical experiments using the MAC scheme show agreement of the numerical results with theoretical analysis.     

Feasibility and modelling for convergence of optical-wireless network – A review

Now a days, the convergence of wireless and optical network gaining popularity to enhance the data carrying capacity, QoS and to support future multimedia applications. The queue theory plays an important role for analytically evaluating such converged networks. In the everyday state of affairs, queue phenomena are observed in the several situations like the health clinic, railway reservation system, automobile laundry centre, bank cash counter, call centre, etc. Similarly, the phenomenon of congestion is ascertained in the telephone network, computer network, communication systems and the Internet wherever nonstop enhancements need to be created to stay up with the quickly growing demand. The subsistence of these adverse clogging effects has lead toward the improvement of queuing theory-based statistical models. Therefore, these queues based statistical models answering to optimising queries corresponding to the way for assigning resources in converged optical and wireless networks. During this paper, a review on the role of the queuing model in wireless & optical communication is presented that concentrate on the present and future aspects of research.     

Multiresponse Estimation With Special Application to Linear Systems of Differential Equations

We present a multiresponse estimation procedure for parameters of systems described by ? = As. The procedure features a generalized Gauss–Newton algorithm for optimizing the determinant criterion, efficient evaluation of the expectation function and its derivatives directly from the reaction network or compartment diagram, automatic determination of starting values, and eficient computational procedures for handling linear constraints on the responses.     

Optimization by record dynamics

Large dynamical changes in thermalizing glassy systems are triggered by trajectories crossing record sized barriers, a behavior revealing the presence of a hierarchical structure in configuration space. The observation is here turned into a novel local search optimization algorithm dubbed record dynamics optimization, or RDO. RDO uses the Metropolis rule to accept or reject candidate solutions depending on the value of a parameter akin to the temperature and minimizes the cost function of the problem at hand through cycles where its ‘temperature’ is raised and subsequently decreased in order to expediently generate record high (and low) values of the cost function. Below, RDO is introduced and then tested by searching for the ground state of the Edwards–Anderson spin-glass model, in two and three spatial dimensions. A popular and highly efficient optimization algorithm, parallel tempering (PT), is applied to the same problem as a benchmark. RDO and PT turn out to produce solutions of similar quality for similar numerical effort, but RDO is simpler to program and additionally yields geometrical information on the system’s configuration space which is of interest in many applications. In particular, the effectiveness of RDO strongly indicates the presence of the above mentioned hierarchically organized configuration space, with metastable regions indexed by the cost (or energy) of the transition states connecting them.     

On the convergence of iterative methods for stabilized saddle point problems

The convergence of the inexact Uzawa method for stabilized saddle point problems was analysed in a recent paper by Cao, Evans and Qin. We show that this method converges under conditions weaker than those stated in their paper.