二样本问题

For the mapping problems in binary field, a learning algorthm is given to choose better hidden nodes for multilayer feedforward networks, by means of solving maximum separated groups.(7) Expert system for evaluating oil shale is constructed based on neural networks.

改进了前述两种多层前向网络学习算法，使网络的隐节点减少一个；对二值域上的映射问题，通过求解选择样本集极大划分组的手法，给出了一种选择较优隐节点的多层前向网络学习算法。

And then fuzzy least square support vector machines are proposed based on support vector domain description. Data samples in the feature space are described and the smallest enclosing hypersphere is obtained. The fuzzy membership value to each sample point is determined according to the distance of each sample from the center of the hypersphere, which can reduce the effect of outliers.

该方法先对样本进行数据域描述得到一个包含该组数据的最小半径的超球，再根据特征空间中样本与超球球心的距离确定它们的隶属度，减少了奇异点的影响；把所要求解的约束凸二次优化问题转化为正定线性方程组，并采用快速Cholesky分解的方法求解该方程组。

Constructed by standard binary classes support vector machine, present multiclass SVMs are usually very slow to be trained. When a large number of categories of data are to be classified, the training work could be very difficult. By extending the hypersphere one-class SVM to a hypersphere multiclass SVM, we build a fast training classifier HSOC-SVM. Its training speed is higher than that of the present multiclass classifiers, because each category data trains only one HSOC-SVM.

目前的多类分类器大多是经二分类器组合而成的，存在训练速度较慢的问题，在分类类别多的时候，会遇到很大困难，超球体多类支持向量机将超球体单类支持向量机扩展到多类问题，由于每类样本只参与一个超球体支持向量机的训练，因此，这是一种直接多类分类器，训练效率明显提高。

In view of the existing problems of traditional resource allocating networks, a design method for RAN based on rough set and orthogonal least square was proposed. Firstly, rough set was applied to intelligent data analysis for extracting typical characteristics from the training samples, and then OLS was used to select best centers as the hidden layer nodes.

针对常用的资源分配网络存在的问题，提出了一种基于粗糙集和正交最小二乘的资源分配网络设计方法，通过粗糙集数据分析与处理提取训练样本中典型的数据特征，再结合正交最小二乘学习算法选取对输出能量贡献最大的数据中心加入到隐层节点。

In view of the existing problems of traditional resource allocating networks, a design method for RAN based on rough set and orthogonal least square was proposed. Firstly, rough set was applied to intelligent data analysis for extracting typical characteristics from the training samples, and then OLS was used to select best renters as the hidden layer nodes.

针对常用的资源分配网络存在的问题，提出了一种基于粗糙集和正交最小二乘的资源分配网络设计方法，通过粗糙集数据分析与处理提取训练样本中典型的数据特征，再结合正交最小二乘学习算法选取对输出能量贡献最大的数据中心加入到隐层节点。

In view of disadvantages in present defect reconstruction methods,such as over-long time or excessive need for training samples,a new defect reconstruction method based on similar model and genetic algorithm was investigated as follows:similar model between non-axisymetric defect and axisymetric defect was represented,and proportionality factor was deduced by skin depth equation.

针对以往缺陷重构方法中普遍存在的计算时间长或需要大量训练样本的缺点，提出一种基于相似模型和遗传算法的缺陷快速重构方法，分析非轴对称缺陷与轴对称缺陷之间存在的相似模型，并利用集肤深度公式推导出两者之间的缩比因子，从而将耗时巨大的缺陷重构问题转换为二维轴对称情况下的计算问题；同时为进一步缩短重构时间，提出一种加快遗传算法收敛速度的方法。

The measurement system error model analysis is introduced into power system real time network state analysis as an important part for the first time, which can improve on the state estimation quality and provide the capability to monitor the operation of the measurement system; 2. The theory and algorithm of the on-line estimation and update of measurement noise variance based on the relation between the residual variance and noise variance. The statistic properties of the sample variance are discussed and the relation between the estimation precision and sample size under given confidence level is derived; 3. The theory and algorithm of detection and identification of measurement bias are presented, which is based on the relation between residual mean and noise mean. The statistic properties of sample mean are discussed and the relation between estimation precision and sample size is derived; 4. The Givens orthogonal transformation algorithm is selected to be the essential algorithm of state estimation, the fast orthogonal transformation algorithm with damp factor and the algorithm which can handle the zero injection measurements efficiently are presented; 5. The quantity analysis theory of bad data detectivity and identifibility are presented, which describes the relation between the elements in matrix W〓 and bad data amplitude and can provide the theory base for measurement system design and valuation.

一、首次将量测系统误差模型分析做为一个环节引入电力系统实时网络状态分析中，为EMS系统增加了实时监视系统运行、修正量测系统误差模型的新功能，进一步发挥了实时网络状态分析应用软件的潜力；二、首次提出了应用样本方差在线估计与修正量测系统误差方差的基本理论，讨论了样本方差的统计性质和概率分布，推导出了样本容量、估计精度和置信度之间的关系，给出了在线估计与修正量测系统误差方差的算法；三、首次提出了应用样本均值在线检测与辨识量测偏差的基本理论，讨论了样本均值的统计性质，推导了样本容量、估计精度和置信度之间的基本关系，给出了在线检测与辨识量测偏差的算法；四、在状态估计算法设计中，以Givens变换算法做为基本算法，提出了快速正交变换阻尼因子法和可以有效地处理零注入量测的混合法，并对实时应用中的一些问题进行了讨论；五、提出了不良数据可检测性与可辨识性的定量分析理论，揭示了描述量测系统配置、网络结构与参数的残差灵敏度矩阵中的元素与不良数据的幅值在可检测、可辨识能力上的定量关系，为量测系统配置设计与评价提供了理论基础；六、综合国内外最新研究成果，采用自适应自回归预测技术和稀疏矢量技术，构造了较完善的不良数据检测与辨识算法。

Hierarchical linear modeling is the most frequently used technique for analyzing multilevel data. In this paper, the twelve key issues in multilevel research and/or using HLM software were reviewed, and provided their exact meanings and feasible solutions. The 12 key issues included intraclass correlation coefficient, sample sizes, formation of organizational constructs, centering problems, fixed effect and random effect, estimation methods, goodness of fit index, explained variation, multicollinearity, robust standard error, level one variance equation, and empirical Bayes estimates.

本文针对阶层线性模式在多层次研究上有关於技术、测量与方法论所遭遇的问题，进行文献整理并讨论其实务意涵，并依据使用时机、抽样议题、资料汇总、分析方法、与模式评估与估算等五方面整理出十二项重要议题与解决之道，包括：组内相关系数的意义与判断准则、多层次研究各层的样本数问题、变数聚合成组织层次问题、中心化的意义、固定效果与随机效果的设定、估计法的选择、适配度的比较、解释变异量的计算、多元共线性的分析、强韧性标准误、第一层误差项变异数与实证贝氏估计值的运用。

As a result, it is convenient for the recognition system to use the simple linear classifier to segment the sample space, so that the accuracy and distortion invariant ability can be easily achieved. The classification structure of a bifurcating tree is utilized to decompose the classification problem of multi-classes to the multi-stage classification of two classes, which simplifies the design of the classifier, and improves the searching efficiency as well. A gray-scale analysis based method for extracting features from optical correlation result is proposed.

提出了二叉树鉴别分类方法，在使用同类图象的多个典型畸变样本训练识别系统的基础上，利用鉴别分析方法优化各类图象的特征向量在样本空间中的分布，不仅有利于同时提高系统的准确性和抗畸变能力，而且便于采用简单的线性分类器来划分样本空间；利用二叉树分类结构，将多类样本的分类问题分解为多级二类样本的分类问题，不仅简化了分类器的设计，而且提高了搜索效率。

A kind of quatratic programming technique,Potential Reduction Algorithm,is used for solving the optimization problem.

本文利用二次规划方法，讨论对神经网络训练样本的吸引半径的优化问题，并借用二次规划中的PRA算法求该优化解，得到一种新的神经网络基于规划的学习算法。

And Pharaoh spoke to Joseph, saying, Your father and your brothers have come to you.

47：5 法老对约瑟说，你父亲和你弟兄们到你这里来了。

Additionally, the approximate flattening of surface strip using lines linking midpoints on perpendicular lines between geodesic curves and the unconditional extreme value method are discussed.

提出了用测地线方程、曲面上两点间短程线来计算膜结构曲面测地线的方法，同时，采用测地线间垂线的中点连线和用无约束极值法进行空间条状曲面近似展开的分析。