离散分布

Application of discrete-ordinate method in calculation of light angular distribution in biological tissue［J］.

离散坐标法在计算生物组织内光场空间角分布中的应用［J］。

Algorithm application of neural networks.Ⅲ Implementation of neuro-computers. The main contribution of the dissertation can be summarized as follows: 1 Hopf bifurcation of three kind of neural networks are discussed in detail, including type of discrete time delay, type of time delay with weak kernel and strong kernel as well as the proof of existence of bifurcation. Other problems such as asymptotic stability of bifurcation periodic solution, algorithm of determining the bifurcation direction, asymptotic stability and style of periodic solution are also studied. The average time delay is chosen as the bifurcation parameter, phenomena pertinent to system states of the continuous time delay network with strong kernel evolving from stable to oscillating, then back to stable again are observed.

论文的主要创新之处可以归纳如下： 1）针对目前国内外对神经网络的分岔研究较少的情况，论文中详细讨论了带离散时延神经网络、带弱核的连续时延神经网络、带连续分布时延且具有强核的神经网络的Hopf分岔现象，从理论上证明了Hopf分岔的存在性，并研究了分岔周期解的渐近稳定性，得到了确定周期解的渐近稳定性、分岔方向、周期解的渐近形式的算法；用平均时延作为分岔参数，发现带强核的连续时延神经网络中存在着系统的状态由稳定变化到振荡现象，当继续增加平均时延参数时，又从振荡变为稳定这一特殊的动力学现象。

This dissertation shows each unstable asymptotic equilibrium point is just the pattern that refuses to be recognized by combining the idea about constraint and projection with the continuous network LSSM and the discrete network BSB, eliminates some shortcomings of LSSM and BSB in the distribution of attraction domain. LSSM and BSB are then improved to be ideal associative memory and can be implemented by circuit.

将约束投影思想与连续神经网络LSSM和离散神经网络BSB相结合，说明了不渐近稳定的平衡点恰是实际中拒识模式，根除了LSSM和BSB在吸引域分布等方面的一些缺点，将它们完善成为理想的联想记忆器，并且它可用电路实现。

Coefficient distribution of the APIDCSF is similar to that of circular quincunx Neville filter and its interpolating performance outperforms the Neville filters with high orders vanishing moments.The aim of bidimensional interpolating is to diminish the error between the true value and the estimated one.

二维内插的质量是以预测值与真值之间的误差来衡量的，按照最小均方误差的准则，利用二维自相关函数模型和环形内插滤波器的系数分布特点构造了最小均方误差意义上的环形最优内插器，实验证明了全相位离散反余弦内插器近似于最优内插器。

According to the principle of profit maximization and based on a discrete demand function which can be represented as a negative binomial distribution,the optimal pricing model for the perishable products is established.

基于一种负二项分布的离散需求函数，并在利润最大化原则下，建立了时效商品最优定价模型。

Binomial distribution is preferred in testability demonstration,but it is discrete.Because of calculation difficulty,the sampling plan is usually worked out with some adaptations.

在测试性验证中首选的二项分布是离散取值的，由于计算困难，确定抽样方案通常采用一些变通的方法。

In the fifth and the sixth chapter, a class of interval BAM neural networks withmixed delays under uncertainty and a class of stochastic interval BAM neural networkswith mixed delays are introduced and studied, which include many well-known neuralnetworks as special cases.

第五章和第六章分别介绍了一类具有不确定性的带混合时滞的区间BAM神经网络和一类带混合时滞的区间随机BAM神经网络，这里混合时滞意味着同时出现离散和分布时滞。

Traditional method to coarsen and disperse the spatial data to lower administer unit in order to let spatial data be agreement to statistic data loses some spatial information. The comprehensive indexes got from these data of course cannot express the land-use spatial differentiation regularity well.

传统的研究方法是将空间数据"粗化"和"离散化"到行政单元上，与统计数据匹配，结果失去了一些指标的空间分布信息，得到的综合评价指数并不能较好地反映区域土地利用的空间分异规律。

Uncertain general MDP with compact parameter set is discussed, and a sub-optimal adaptive control method is proposed.

分析和证明了一般MDP的极限分布和目标函数的连续性，并在此基础上将参数集离散化方法推广到一般MDP，获得了能够以任意给定精度实现次优的自适应决策方法。

Methods for feature space discretization based cluster algorithm and class conditional probability distribution estimation are proposed.

首先针对应用贝叶斯决策理论时难以估计类条件概率密度函数的问题，提出了特征空间离散化及相应的类条件概率分布的估计方法。

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。