算法

The cause of the target fade-out in Neural Network nonuniformity correction scheme for infrared focal plane arraywas studied. Based on analyzing the strength and limitation of edge-directed NN scheme and nonuniformity correction combining one-point calibration and NN-NUC, a new combinational algorithm for nonuniformity correction was proposed.

中文摘要：分析了传统神经网络非均匀性校正算法在空域处理过程中产生目标退化的原因，在总结基于边缘指导的神经网络校正算法与一点定标和神经网络结合的校正算法的基础上，提出了新的组合校正算法。

Analysis for testability is widely used in DFT and test pattern generation, but most testability analysis algorithms have the shortcoming of either inaccuracy or being too complicated. A new method to calculate the signal probabilities of the nodes in IC is pre-sented in Chapter 7 of this dissertation. The signal probabilities by our method is more exact than by the previous ones. A symbol of the signal probability at the fan-out node is created and is propagated to the RFON (Reconvergent Fan-out Nodes). At the RFON, the expression containing the symbol is simplified using the rule which corresponds to the law of identity in Boolean algebra. After the expression is simplified, the symbol is substituted with its numeric value. Theoretical analysis shows that this method is concise and has a low computing complexity.

可测试性分析在可测试性设计和测试生成中都有着广泛的应用，然而大部分的可测性分析方法都存在不精确或者过于复杂的毛病，在第七章中作者提出了一种新的计算IC电路节点的信号概率值的符号算法，该算法得出的信号概率值比用以往的算法法得出的要更准确，也更简洁，针对电路中的扇出节点重汇聚现象，我们在每一个扇出节点定义了一个信号概率的符号，在以后的计算中一直使用该符号参与计算，直到该符号传播至重汇聚处，利用一定的规则把含符号的表达式化简，再将符号的真实值代入，在计算符号概率的过程中定义了一个和布尔代数中的同一律相对应的运算规则，通过该运算规则，使得信号概率的结果更加准确，符号算法具有简洁方便，计算复杂度小的特点 4。

We improved the merit function in quasi-feasible interior point method and used the exponential function that is the more general function to prove the existence of one order optimality point in feasible problem. We discussed the inner and outer algorithms and made the convergence theorem.

在算法中改进了拟可行内点法中的值函数，使用指数形式的更一般函数，用此值函数证明了可行性问题的一阶最优性点的存在性，并通过对内部算法及外部算法的讨论得到了算法的收敛性定理。

For its simple computational complexity, fast convergence speed and many other advantages, adaptive filer has been widely used.

分别就几种算法在算法原理，算法性能分析和计算机仿真等方面来说明各种算法的优越性。

In order to preserve more details while erasing mixed noise, a new algorithm based on the structure of modified trimmed mean filter was proposed. Characters of the new one are: 1 replaces the median filter with center weighted median which shows better performance in preserving details; 2 uses the filtering result to estimate the deviation of Gaussian noise. The estimating result is used to change the filter parameter, so the new algorithm is robust.

为了在滤除退化图像中混合噪声的同时能更好地保护图像的细节，文中提出了基于改进均值滤波结构的新算法，其特点为： 1）用中心加权中值滤波算法替代MTM 方法中的中值滤波算法以更好地保护细节； 2）提出了一种直接利用滤波结果来估计混合噪声中高斯噪声方差的方法，估计的结果被用于滤波器系数的调整，使得新算法有很好的鲁棒性。

The theory and the implementation of the genetic algorithms are discussed in detail. The question on how to choose the crossover probability, the mutation probability, the scale of population and the numbers of the generation is discussed. Then, the mathematics model of the optimal design is established. Exerting the finite element of elastic base beam and using FORTRAN to write the GAFORTRAN program. The program includes the common genetic algorithms and the modified genetic algorithms.

详细介绍了遗传算法的理论和实现技术，探讨了交叉概率、变异概率、群体规模、进化代数等变量的选取问题，建立起了基于遗传算法的深基坑支护结构设计的优化模型，结合弹性地基梁有限元法，利用FORTRAN语言编制了GAFORTRAN优化程序，程序中包括普通遗传算法和改进遗传算法。

Digital receiver and its forepart operational principle are discussed, then me...

本文讨论了数字接收机以及前端的工作过程，然后详细阐述了瞬时自相关、短时傅立叶变换、WD算法、过零检测和小波分析等各种算法在脉内调制分析上的应用，分析了各算法的优缺点，提出一种并行处理模式：即针对不同的信号采用不同的算法对脉内调制信号进行分析。

On the basis of partially observable Markov decision processes,two finite-memory policy-gradient algorithms,that is,model-based GAMP algorithm and model-free IState-GPOMDP algorithm,were implemented,and employed in the simulation of a robot walking in a maze.

在部分可观Markov决策过程的基础上，实现了两个有限记忆的策略梯度优化算法：基于模型的GAMP算法和无模型的IState-GPOMDP算法，并利用该算法对机器人走迷宫的问题进行了仿真。

The dissertation focuses on performance gradient estimation and SRO algorithm of DES aswell as other related topics,such as modeling of DES and discrete event simulation,performancemeasure Hessian matrix estimation,Stochastic Approximationand so on.General Semi-Markov Processis adapted to describe behaviour of DES and discrete eventsimulation.A general result is obtained and applied to analyze strong consistency of performancegradient Perturbation Analysisestimator.Conditions for effectiveness of InfinitesimalPerturbation Analysisare also discussed.New PA algorithms,such as server PA,PA usingDirac generalized function,job sequence PA etc.,are proposed to deal with discreteparameters.Furthermore,strategies of combining PA algorithms with certain iterative rules areanalyzed widely in order to optimize DES parameters and some SRO algorithms are constructed toslove DES optimization problems.

论文将DES性能分析领域中性能梯度估计和单运行仿真优化作为主要的研究对象，探讨了与性能梯度估计有关的DES和离散事件仿真过程建模、高阶导数估计以及随机逼近等理论和方法；建立了具有一般性的DES和离散事件仿真的广义半Markov过程模型；获得了分析性能梯度估计量强相合性的理论结果和无穷小摄动分析算法应用的条件；提出了诸如服务台数目的摄动分析、奇异摄动分析、串联生产线和加工车间中工件加工序的摄动分析等一系列摄动的思想和算法，对连续参数和离散参数都进行了深入的研究；对摄动分析算法同优化规则结合实现仿真优化的策略也进行了广泛的分析，构造了新型的变界截尾的随机逼近算法，并用于DES的单运行仿真优化研究；最后，给出了DES性能分析方法的实证性研究结果。

According to the simulation testing and the comparision with several algorithm was verified and the global convergence property of the algorithm was proved.

通过仿真算法及与几种算法的比较结果验证了算法的有效性，证明了算法的全局收敛性。

The split between the two groups can hardly be papered over.

这两个团体间的分歧难以掩饰。

This approach not only encourages a greater number of responses, but minimizes the likelihood of stale groupthink.

这种做法不仅鼓励了更多的反应，而且减少跟风的可能性。