收敛性的

In order to improve the BP Arithmetic's study rate, we do some improvement, and get more accurate results. Then we introduce the improved GA which used to improve the BP Arithmetic. Another DM technology--Fuzzy theory which will be used in the Pre-prepare of the load and relative data.

接着用改进后的遗传算法结合改进后的BP算法来改善神经网络的局部收敛性，同时采用了模糊理论的知识，对各种数据进行归一化和修正处理，提高了网络的预测精度和学习效率。

MRF-MBNN is also robust to different preassigned class number, the network will convergence to nearly same results in few iterations. The segmentation results for chromosome images is far better than those in the thesis are. It is same for Lena image.

MRF-MBNN网络除了具有MBNN网络对预定类别数的强鲁棒性和对不同类别数网络收敛快的特点外，将之用于染色体图像的分割效果远优于本文中所述的其他方法，而对Lena图像也获得了优于MBNN的结果。

In Chapter 5, the authors study the convergence properties of the gradient projection method for the constrained optimization problem. In this chapter, a new step-size rule, which avoids fulfiling the classical line search and includes choosing a constant as the step size as a special case, is presented and analyzed.

第五章研究了求解约束最优化问题的梯度投影方法，在步长的选取时采用了一种新的策略，这种策略不需要进行传统的线搜索且包含步长取常数这种特例，在较弱的条件下，证明了梯度投影方法的全局收敛性。

By using generalized projection matrix, conditions are given on the scalars in the three term-memory gradient direction to ensure that the three term-memory gradient projection direction is a descent direction. A new three term-memory gradient projection method for nonlinear programming with nonlinear equality and in-equality constraints is presented.

利用广义投影矩阵，对求解无约束规划的三项记忆梯度算法中的参数给一条件，确定它们的取值范围，以保证得到目标函数的三项记忆梯度广义投影下降方向，建立了求解非线性等式和不等式约束优化问题的三项记忆梯度广义投影算法，并证明了算法的收敛性。

First of all, on the base of the reduced gradient method of Wolfe, we added accurate one dimension searching method into it and gained a new algorithm (algorithm 1,2) for the nonlinear programming with linear equality constraints and quadratic programming problems, and proved convergence of algorithm, then, we use above algo...

首先本文在Wolfe既约梯度法的基础上，针对具有线性等式约束的非线性规划问题和二次规划问题，引入了精确的一维搜索，得到了带一维搜索的新算法(算法1、2)，并给出了算法的收敛性证明。

When the sojourn time of deteriorating state, inspection interval and maintenance time are the stochastic variables which follow the general distribution, the phase-type approach is proposed to preserve the analytical tractability of the transition probability matrix. To solve the problem that the PH distribution changes the state space of system, the value iteration algorithm for the semi-Markov decision process is improved to get the optimal inspection and maintenance policy.

针对模型中劣化阶段停留时间满足一般分布的情况，提出了一类基于位相型分布拟合的近似处理方法，将物理意义上的状态进行概率扩展，简化了维修决策的求解过程，并给出了对应的改进迭代算法，同时从理论上证明了该算法的收敛性。

Based on the research of some recent algorithms work of the initial solution estimation of inverse simulation approach,an improved algorithm,named projecting and averaging linear deformation method,is presented,which enables the inverse approach more stable and improves its convergence greatly.

初始解的确定是板料冲压成形有限元反向分析中的一个重要环节，初始解的好坏会影响后续计算的收敛性以及分析计算的效率。

At last, the application of above modeling and control strategies has been demonstrated by the simulations of a nonlinear dc motor control problem. Under this control strategy, system identification phase is not needed, and only a controller based on the recurrent neural network is provided. Because the control and learning laws are designed with stable theory, stability and convergence problems are solved here with rigorous analysis, which can not been analyzed easily when usual BP neural networks are used.

最后通过非线性电机模型的仿真研究，表明了基于上述动态动态递归神经网络模型对非线性系统进行自适应控制的实际可行性，并且采用这种方法不需要对系统进行离线辨识，整个系统只采用了一个基于动态递归神经网络的控制器，由于整个系统是基于稳定性理论设计的，有效地解决了应用BP网络等其它前向网络时系统的稳定性和收敛性难以分析的问题。

Although the unstable manifold of the fast subsystem is different from that of the full system, the stable manifold of the fast subsystem is almost the same as that of the full system, so the domain of attractor of the full system can be given by that of the fast subsystem.

除了理论分析外，同时也给出了数值模拟的结果，包括分支图，轨线，吸引域，各变量的收敛性等。本章给出的稳定平衡点的吸引域对电力系统实行适时控制起到很重要的作用。

We study the attractive manifold method of the iterative learning control for a class of degenerate systems in the chapter Ⅲ. When the learning control systems don't satisfy the convergence condition, we present the learning control method satisfying the condition by constructing a stable and attractive manifold corresponding system. This method makes the system with better robustness.

第三章主要研究了一类退化系统目标跟踪的迭代学习控制的吸引流形方法，在系统不满足一般收敛条件的情况下，我们通过构造一个相应于所给系统稳定而吸引的流形，导出所满足要求的迭代学习控制，该方法对系统有较好的鲁棒性。

This one mode pays close attention to network credence foundation of the businessman very much.

这一模式非常关注商人的网络信用基础。

Cell morphology of bacterial ghost of Pasteurella multocida was observed by scanning electron microscopy and inactivation ratio was estimated by CFU analysi.

扫描电镜观察多杀性巴氏杆菌细菌幽灵和菌落形成单位评价遗传灭活率。