基于案例学习的排球机器人运动规划及其支持向量回归实现

针对混合型控制问题,以排球任务为例研究机器人的运动规划.模拟人类球员通过经验积累而采取相应动作的行为学习模式,采取案例学习的方式解决球的初始状态微小变化(仅发球速度和角度变化)时的运动规划问题.由于支持向量回归(SVR)在处理小样本问题的优越性并受局部学习思想的启发,采用局部加权SVR(LW-SVR)实现案例学习.结果证明,LW-SVR的学习精度较RBF神经网络和SVR明显提高.     

非参数计量经济模型的变窗宽核估计

在随机设计(模型中所有变量为随机变量)下,提出了非参数计量经济模型的变窗宽核估计,并利用概率论中大数定理和中心极限定理,在内点处证明了它的一致性和渐近正态性.它在内点处的收敛速度达到了非参数函数估计的最优收敛速度.     

特征筛选对脑-机接口信号单次提取精度的影响

脑-之机接口的核心问题之一是通信载体信号的单次提取．在构建脑控拼写器的过程中，通过“模拟自然阅读”诱发模式产生的视觉诱发电位作为人脑与计算机之间的通信载体，采用支持向量机方法进行特征信号的单次识别．为提高识别精度，详细研究了信号时程、时段的选择对模式识别精度的影响．结果表明，信号时程越长分类精度越高，时程达到300ms时，分类精度就可达到最大值（且趋于饱和）；信号时段的选择对分类精度亦有较大影响，最佳时段在靶刺激出现后约250～350ms作为起始处．这一结果为提高系统的整体速度与精度打下了基础．     

Mining the Change of the Quality of Stored Grain

The quality of grain changes continually during its storage, including the change of its physical characteristics and physiological characteristics. This paper presents an approach to predict the change of the quality of stored grain with data mining technology. Logistic Regression, Decision Tree and Muitilayer Perceptron are applied to predict the change of the grains＇ quality control index and to obtain the grains＇ quality change probability. The grain sampling with higher probability can be processed earlier.     

缺失数据下局部M-估计的相合性和渐近正态性

在缺失响应变量的不完全数据下,对非参数回归模型进行研究,利用稳健的变窗宽局部线性回归的方法,给出了回归函数m(x)的估计,并证明了局部M-估计具有相合性和渐近正态性.所提出的方法继承了局部多项式回归的优点并且克服了最小二乘方法缺乏稳健性的缺点.并且使用变窗宽提高了所得M-估计的可塑性,使之能成功地处理空间非齐次曲线、异方差性及非均匀设计密度.所得估计的渐近结果为求渐近最优方案以及直接从数据估计最优变窗宽提供了理论基础.     

利用零矢量矩阵法研究正交曲线坐标系中的Galileo变换

本文中引入零矢量矩阵概念，利用零矢量矩阵法提出Galileo变换下正交曲线坐标系中质点的坐标，速度和加速度的矩阵形式的公式，而且推导Galileo变换下柱座标系和球座标系中质点的速度和加速度的矩阵形式．     

Response of downstream sediment to water resource development in mainstream of the Lancang River

The impact of hydropower cascades of the Lancang River on downstream sedimentation has become one of the key problems in the research of the transboundary water resource development and its possible environmental effects. In this paper, the newest sediment data series spanning 17 years （1987-2003） at Yunjinghong hydrometric station and Chiang Saen hydrometric station, in northern Thailand, are analyzed to reveal these effects. Monthly mean sediment concentration data of the two stations are compared in-phase and yearly series are analyzed by the method of correlation and regression analysis and causality test. The analysis result shows that, though both of them are located at lower reaches of Manwan and Dachaoshan hydropower stations which are already constructed and in use, the sediment responses of the two stations are different under the regulation of the hydropower stations Annual series at Yunjinghong station are positively correlated but do not match well with those at Chiang Saen station. In addition, the results of Granger causality test on the time series also suggest that the series of maximum monthly mean sediment concentration of the Yunjinghong station each year may give cause for the ones of Chiang Saen station, while the series of minimum monthly mean sediment concentration each year and annual mean sediment concentration of Yunjinghong do not evidently give cause for the ones of the other station. Undoubtedly, this discovery will provide new evidences to assess the impacts of cascades construction in mainstream of the Lancang River on the variation of downstream sediment and its transboundary effects.     

一种基于KNN与改进SVM的车牌字符识别算法

提出了一种将KNN(K近邻)和支持向量机相结合的字符识别算法.首先用KNN对字符进行判断,如果输出的置信度大于阈值,则认为分类正确;如果小于阈值则采用支持向量机进行判决.改进了SVM分类器,通过调整支持向量机的分类超平面改进了支持向量机的性能.将算法应用到实际的车牌字符识别中,识别结果表明,这种方法在提高识别速度的同时,有效提高了字符的识别精度.     

关于改进的支持向量机的研究

支持向量机是一种基于统计学习理论和对偶理论的分类和函数估计方法,其推广能力强和全局最优的特点引起了学者的广泛关注。本论文对目前所提出的各类改进的支持向量机进行了初步的研究和分析。