《深入浅出数据分析》以类似“章回小说”的活泼形式，生动地向读者展现优秀的数据分析人员应知应会的技术：数据分析基本步骤、实验方法、最优化方法、假设检验方法、贝叶斯统计方法、主观概率法、启发法、直方图法、回归法、误差处理、相关数据库、数据整理技巧；正文以后，意犹未尽地以三篇附录介绍数据分析十大要务、R工具及ToolPak工具，在充分展现《深入浅出数据分析》目标知识以外，为读者搭建了走向下一步深入研究的桥梁。
《深入浅出数据分析》构思跌宕起伏，行文妙趣横生，无论读者是职场老手，还是业界新人；无论是字斟句酌，还是信手翻阅，都能跟着文字在职场中走上几回，体味数据分析领域的乐趣与挑战。
"Data analysis" similar to "novels" lively form, vividly to readers excellent data analysis personnel should be aware of the data analysis technology: basic steps, experiment method, optimization method and hypothesis testing methods, Bayesian statistics method, subjective probability method, heuristic method, histogram method, regression method, error handling, database, data processing skills; text, enough to three appendix introduces the data analysis of ten major priority, R tools and ToolPak tools in the show "simple data analysis" target knowledge, to build a bridge to the next study for readers.
"Data analysis" concept of ups and downs, the workplace is full of wit and humour, regardless of whether the reader or industry veteran, whether it is new; carefully worded, or Xinshou read, followed by the text in the workplace can go a few times, fun and challenge in the field of data analysis experience. (2018-01-11, PDF, 5167KB, 下载20次)
聚类金字塔树-一种新的高维空间数据索引方法
提出了一种新的有效的高维空间数据索引方法 聚类金字塔树.它先对不均匀分布数据进行聚类 处理,然后对聚类的结果实施金字塔分割和存储,由此建立一种有效的索引结构.文中给出了聚类金字塔树的几种查询算法.实验证明:处理不均匀分布数据时,聚 类金字塔树无论在页面访问次数,还是在CPU总占用时间上都优于金字塔树.
Pyramid Tree Cluster- A new high-dimensional data indexing method presents a new and effective method for high dimensional spatial data indexing- Pyramid clustering tree it first on the uneven distribution of data clustering, and then results clustering implementation pyramid segmentation and storage, thereby establishing an effective index structure paper presents several query clustering algorithm pyramid tree proved: when dealing with the uneven distribution of data, both in the clustering tree pyramid page visits, or in the total of occupied CPU time is better than the pyramid tree. (2017-03-05, PDF, 252KB, 下载2次)
针对采用回声状态网络预测多元混沌时间序列时存在的病态解问题 , 本文建立了因子回声状态网络模型 , 通过因子分析 (Factor analysis, FA) 方法提取高维储备池状态矩阵的公因子 , 去除冗余和噪声成分 .
When an echo state network is used to predict multivariate time series, there may exist ill-posed problem. This pa-
per proposes a novel prediction model, named factor echo state
network, to solve the problem. It uses a factor analysis (FA) algorithm to extract the common factors of the reservoir matrix,and to remove the redundancies and noises. (2016-06-12, PDF, 752KB, 下载14次)
张量是一个定义在一些向量空间和一些对偶空间的笛卡儿积上的多线性函数，其坐标在 维空间内，有 个分量的一种量， 其中每个分量都是坐标的函数， 而在坐标变换时，这些分量也依照某些规则作线性变换。 称为该张量的秩或阶（与矩阵的秩和阶均无关系）。
Is a definition of a Cartesian tensor in some vector space and some dual space product multi linear function of the coordinates in the dimensional space, there is a quantity of components, where each component is function of coordinates, and the coordinate transformation, these components are in accordance with certain rules for linear transform. The rank order of the tensor is not related to the rank order of the matrix. (2015-12-14, PDF, 60KB, 下载4次)
主成分分析是把多个特征映射为少数几个综合特征的一种统计分析方法。在多特征的研究中，往往由于特征个数太多，且彼此之间存在着一定的相关性，因而使得所观测的数据在一定程度上有信息的重叠。当特征较多时，在高维空间中研究样本的分布规律就更麻烦。主成分分析采取一种降维的方法，找出几个综合因子来代表原来众多的特征，使这些综合因子尽可能地反映原来变量的信息，而且彼此之间互不相关，从而达到简化的目的。主成分的表示相当于把原来的特征进行坐标变换（乘以一个变换矩阵），得到相关性较小（严格来说是零）的综合因子。
The principal component analysis of multiple feature maps for a few characteristics of a statistical analysis method. In multi-feature, often because of too many number of features, and between the existence of certain correlation, thus making the information overlap of the observed data in a certain extent. Even more trouble when more of the characteristics, distribution of study sample in a high-dimensional space. Principal component analysis a dimensionality reduction approach to identify several factor to represent many of the original features of the factor as much as possible to reflect the original variable, and unrelated to each other, so as to achieve the purpose of simplified . A representation of the main component (multiplied by a transformation matrix), and obtain a low correlation (strictly speaking, is zero) integrated factor equivalent to the original feature coordinate transformation is performed. (2012-10-29, PDF, 227KB, 下载3次)
该文用不同的本构关系研究了金属衬底手征
涂层的吸波特性,并进行了比较. 结果表明,只要手征媒质的宏观电磁参数统一后,就能使用不同的本构关系得到
相同的结果,进而揭示本构关系的等效性. 数值计算结果表明,在基质中掺加手征体能改变基质的介电常数和磁导
率,从而对反射系数产生影响.
The use of different constitutive relations to study chiral metal coating absorbing characteristics, and compared. The results show that the chiral media as long as the electromagnetic parameters of the macro after the reunification, we can use different constitutive relations have been the same results, which revealed the constitutive relation of equivalence. The numerical results show that in the matrix of mixing matrix chiral physical change in dielectric constant and magnetic permeability, which have an impact on the reflection coefficient. (2007-10-13, PDF, 249KB, 下载16次)