遗传算法在模糊系统优化设计中的应用研究

在模糊系统的变节点自适应模糊神经网络实现的基础上,提出一种混合GA优化算法。该算法采用混合编码策略,利用GA对模糊规则和隶属函数同时优化,而对结论参数则用最小二乘法估计。算法综合了GA强大空间搜索能力和传统优化方法的快速收敛和高精度的优点,在保证全局优化能力的条件下,综合考虑了模糊控制器的复杂程度、训练速度和控制精度。仿真结果及应用表明了该算法的有效性。     

CONVERGENCE PROPERTIES OF MULTI-DIMENSIONAL NEAREST NEIGHBOR DENSITY ESTIMATES FOR THE CASE OF GENERAL KERNEL

In this paper we give weaker conditions to ensure the strong uniform consis-tency of multi-dimensional nearest neighbor (N.N.) estimates with non-uniform kernel andobtain the convergence rates of these estimates on an arbitrary bounded set. The ratescan not be improved in some sense. Obviously, the problem of strong convergence rates ata given point is its special case. The range of applications of estimates is extended.     

Adaptive nonuniformity correction for IRFPA sensors based on neural network framework

For infrared focal plane array sensors, imagery is degraded during signal acquisition, particularly nonuniformity. In this paper, an adaptive nonuniformity correction technique is proposed which simultaneously estimates detector-level and readoutchannel-level correction parameters using neural network approaches. Firstly, an improved neural network framework is designed to compute the desired output. Secondly, an adaptive learning rate rule is used in the gain and offset parameter estimation process. Experimental results show the proposed algorithm can achieve a faster convergence speed and better stability, remove nonuniformity and track parameters drift effectively, and present a good adaptability to scene changes and nonuniformity conditions.     

高阶带阻滤波器优化设计研究

众所周知，传统BP神经网络收敛速度慢、学习效率低。之所以如此，主要原因在于人工神经元输出函数的同一化。本文提出的神经网络模型的主要特点是：用正交基函数作人工神经元的输出函数，而且每个神经元的输出函数各不相同。该神经网络模型有效克服了传统BP神经网络收敛速度慢、学习效率低的致命缺陷。本文还详细研究了FIR线性相位滤波器的幅频特性与余弦基函数神经网络算法的关系，给出了高阶带阻滤波器优化设计实例。计算机仿真结果表明了该算法在高阶带阻滤波器设计中的有效性和优异性能。     

ADAPTIVE SYSTEMS THEORY： SOME BASIC CONCEPTS, METHODS AND RESULTS

The adaptive systems theory to be presented in this paper consists of two closely related parts: adaptive estimation (or filtering, prediction) and adaptive control of dynamical systems. Both adaptive estimation and control are nonlinear mappings of the on-line observed signals of dynamical systems, where the main features are the uncertain-ties in both the system‘s structure and external disturbances, and the non-stationarity and dependency of the system signals. Thus, a key difficulty in establishing a mathematical theory of adaptive systems lies in how to deal with complicated nonlinear stochastic dynamical systems which describe the adaptation processes. In this paper, we will illustrate some of the basic concepts, methods and results through some simple examples. The following fundamental questions will be discussed: How much information is needed for estimation? How to deal with uncertainty by adaptation? How to analyze an adaptive system? What are the convergence or tracking performances of adaptation? How to find the proper rate of adaptation in some sense? We will also explore the following more fundamental questions: How much uncertainty can be dealt with by adaptation ? What are the limitations of adaptation ? How does the performance of adaptation depend on the prior information ? We will partially answer these questions by finding some “critical values“ and establishing some “Impossibility Theorems“ for the capability of adaptation, for several basic classes of nonlinear dynamical control systems with either parametric or nonparametric uncertainties.     

基于模糊推理的自适应BP算法

ＢＰ网络是迄今为止应用最广泛的一种神经网络，但这种算法也存在着收敛速度慢、容易陷入局部极小点等问题．本文在标准ＢＰ算法的基础上提出一种改进ＢＰ算法，称之为自适应ＢＰ算法．这种自适应ＢＰ算法采用模糊规则动态调整学习参数，并且能在学习过程中和学习完成后通过隐节点调整算法优化网络结构，有比标准ＢＰ算法更好的收敛性和更好的泛化能力     

一类自适应松弛算法的算法模型

对于一类自适应松弛算法,本文提出了一个算法模型。在一些通常的假定下,证明了该模型生成的序列的收敛性定理。     

Adaptive tracking of a class of first-order systems with binary-valued observations and fixed thresholds

This paper considers the adaptive tracking problem for a class of first-order systems with binary-valued observations generated via fixed thresholds.A recursive projection algorithm is proposed for parameter estimation based on the statistical properties of the system noise.Then,an adaptive control law is designed via the certainty equivalence principle.By use of the conditional expectations of the innovation and output prediction with respect to the estimates,the closed-loop system is shown to be stable and asymptotically optimal.Meanwhile,the parameter estimate is proved to be both almost surely and mean square convergent,and the convergence rate of the estimation error is also obtained. A numerical example is given to demonstrate the efficiency of the adaptive control law.     

一类自适应随机搜索算法及DEDS优化

采用常规的优化方法难以解决一般ＤＥＤＳ的优化问题，本文提出了一种新的基于仿真的自适应随机搜索技术，用于解决这类问题。实践表明，这种方法具有较高的搜索速度及广泛的适用性。     

AN EXTRAPOLATION METHODFOR FINITE ELEMENT APPROXIMATION OF INTEGRODIFFERENTIAL EQUATIONS WITH PARAMETERS

In this paper,we discuss the accelerating convergence method for finite elementapproximation of integro-differential equations with parameters.As applications,we give theerror estimates of finite element for the first kind of Fredholm integral equation,particularlyfor the Volterra integral equation with kernel condition k(x,x)=0.     

改进的自适应遗传算法在TDOA定位中的应用

针对无线通信到达时间差(time difference of arrival,TDOA)定位技术位置解算为复杂的非线性方程最优化问题,采用实数编码遗传算法,提出了改进的自适应遗传算法。该算法设计了自适应交叉率和变异率的计算公式,考虑了随着进化代数增加种群的整体变化,同时考虑了每代种群不同个体适应度的作用,并引入最优保存策略防止优良个体的破坏,能有效产生新的个体进而摆脱局部最优值的搜索达到全局最优解。仿真结果表明,改进的遗传算法性能稳定,进化收敛速度和TDOA定位估计精度都有较大的提高。     

STOCHASTIC ADAPTIVE CONTROL FOR SYSTEMS WITH NOISE BEING AN ARMA PROCESS

Stochastic adaptive control is considered for the discrete-time multi-input and multi-output system of multi-delay with noise expressed by an ARMA process.The CARIMA model isa special case of the system in question.The optimal adaptive control law is given and it is shown thata quadratic cost function is minimized and the closed-loop system is stable.Further,when the systemis of minimum phase,the convergence rates of parameter estimates and of the cost-function are alsoderived.     

采用梯度平滑措施的自适应符号算法

以自适应回波对消为对象 ,讨论了两种自适应符号算法的梯度估计平滑方法。从理论上对其中一种平滑方案导出的自适应算法的性能进行了详尽的分析。在输入为联合高斯的情况下 ,给出了权值一阶矩和二阶矩的收敛特性公式。结果表明 ,与符号算法相比 ,平滑算法的收敛速度与符号算法基本一致 ,但稳态误差更小 ,自适应步长的取值范围更大。     

ANISOTROPIC BIQUADRATIC ELEMENT WITH SUPERCLOSE RESULT

The main aim of this paper is to study the convergence of biquadratic finite element forthe second order problem on anisotropic meshes.By using some novel approaches and techniques,theoptimal error estimates are obtained.At the same time,the anisotropic superclose results are alsoachieved.Furthermore,the numerical results are given to demonstrate our theoretical analysis.     

高频数据波动率非参数估计及窗宽选择

基于高频数据采用非参数的方法估量波动率,因其能更准确地度量波动率,一直是学者们研究的热点.然而,波动率的所有非参数估计都面临着窗宽的选择问题.由于最优窗宽中往往携带一些难以估计的未知参数,使得在应用过程中确定最优窗宽的具体数值存在困难,从而阻碍了这类估计量的使用.本文以已实现核估计作为波动率非参数估计的代表,构建了一种能自动从实际数据中确定最优窗宽的算法.理论分析的结果表明:算法具有稳定性,其所确定的窗宽是最优窗宽的无偏一致估计量,收敛速度为O(n~(-1/5)).实际数据检验的结果显示,算法是稳定的并且具有良好的适应性,由算法确定的窗宽不依赖初始值的选取.模拟数据的结果显示,相比传统确定窗宽的方法,算法确定的窗宽所对应的估计量具有更高的精度.文中的算法可推广到波动率其他的非参数估计量中,从而为这类估计量的使用铺平道路.     

全变化参数的中位值-极差联合控制图

根据Costa的全变化参数的均值控制图,设计全变化参数中位值-极差联合控制图,记为CVPx-R.将其与静态的中位值-极差联合控制图及其他动态的中位值-极差联合控制图做了比较.数据显示所设计的CVPx-R能较快发现过程的变化.     

带硬时间窗的战场物资配送车辆路径优化

针对战场物资配送中带硬时间窗车辆路径问题的多重模糊性,基于模糊可信性理论建立了多目标模糊期望值模型,提出了一种改进的约束多目标粒子群优化算法。算法采用基于相位空间思想的实数编码方式,提出了带优秀不可行解动态记忆机制的非支配解构造方法,基于自适应栅格和拥挤距离的混合多样性策略维护非支配解集,改进了个体向导更新方式,提高了算法的收敛性能,同时引入局部搜索和变异算子避免算法早熟。仿真实验表明了模型的合理性和算法的有效性。     

STOCHASTIC SYSTEM IDENTIFICATION VIA ADAPTIVE SPECTRAL FACTORIZATION

We consider in this paper the problem of recursive identification for stochastic systemswhen the noise model does not satisfy the positive real condition associated with convergence ofstandard algorithms.To avoid the positive real condition,adaptive spectral factorization techniquesare exploited on the basis of a class of non-standard time-varying recursive Riccati equations.Theasymptotic properties of the Riccati equations are studied as a crucial step to the convergence resultsof the paper.     

The Learning Control and Learning Adaptive Control of General Nonlinear Systems：MIMO Case

ＴｈｅＬｅａｒｎｉｎｇＣｏｎｔｒｏｌａｎｄＬｅａｒｎｉｎｇＡｄａｐｔｉｖｅＣｏｎｔｒｏｌｏｆＧｅｎｅｒａｌＮｏｎｌｉｎｅａｒＳｙｓｔｅｍｓ：ＭＩＭＯＣａｓｅＨＯＵＺｈｏｎｇｓｈｅｎｇａｎｄＨＡＮＺｈｉｇａｎｇ（ＩｎｓｔｉｔｕｔｅｏｆＡｐｐｌｉｅｄＭａ...     

A CLASS OF REVISED BROYDEN ALGORITHMS

In this paper, we discuss the convergence of the Broyden algorithms with revised search direction. Under some inexact llne searches, we prove that the algorithms are globally convergent for continuously differentiable functions and the rate of convergence of the algorithms is one-step superlinear and n-step second-order for uniformly convex objective functions.