diff --git a/docs/14-Unsupervised-Learning/14.9-Nonlinear-Dimension-Reduction-and-Local-Multidimensional-Scaling.md b/docs/14-Unsupervised-Learning/14.9-Nonlinear-Dimension-Reduction-and-Local-Multidimensional-Scaling.md
index ab294f1be0..36b066a973 100644
--- a/docs/14-Unsupervised-Learning/14.9-Nonlinear-Dimension-Reduction-and-Local-Multidimensional-Scaling.md
+++ b/docs/14-Unsupervised-Learning/14.9-Nonlinear-Dimension-Reduction-and-Local-Multidimensional-Scaling.md
@@ -4,7 +4,7 @@
 | ---- | ---------------------------------------- |
 | 翻译   | szcf-weiya                               |
 | 时间   | 2017-09-04                   |
-|更新| 2018-01-22|
+|更新| 2020-02-20 15:46:34|
 |状态|Done|
 
 最近有人提出了一些用于非线性降维的方法，类似主曲面的思想．想法是将数据看成位于一个嵌在高维空间中的 **固有低维非线性流形 (intrinsically low-dimensional nonlinear manifold)** 的附近．这些方法可以看成是“压扁(flattening)”流形，因此将数据降低至低维坐标系统中，用于表示点在流形中的相对位置．在信噪比非常高的问题中非常有用（比如，物理系统），而且对于有低信噪比的观测数据不是很有用．
@@ -15,14 +15,14 @@
 
 我们将简短地介绍用作非线性降维和流形映射的三个新方法．
 
-**等距特征映射算法(ISOMAP)**（Tenenbaum et al. 2000[^1]）构造了一个图来近似沿着流形的点之间的测地线距离．具体地，对每个数据点，我们找到其邻居——距该点的某个欧式距离范围内的点．我们构造任意两个邻居点间用边相连的图．任意两点的测地线距离则用图中点的最短路径来近似．最终，对图的距离应用经典缩放，来得到低纬映射．
+**等距特征映射算法 (Isometric feature mapping, ISOMAP)** (Tenenbaum et al. 2000[^1]) 构造了一个图来近似沿着流形的点之间的测地线距离．具体地，对每个数据点，我们找到其邻居——距该点的某个欧式距离范围内的点．我们构造任意两个邻居点间用边相连的图．任意两点的测地线距离则用图中点的最短路径来近似．最终，对图的距离应用经典缩放，来得到低维映射．
 
-**局部线性内嵌(LLE)**(Roweis and Saul, 2000[^2])采用完全不同的方式，它试图保持高维数据的局部仿射结构．每个数据点用邻居点的线性组合来近似．于是通过寻找保持局部近似的最优方式构造低维表示．细节非常有趣，所以在这里给出：
+**局部线性内嵌 (Local linear embedding, LLE)**(Roweis and Saul, 2000[^2])采用完全不同的方式，它试图保持高维数据的局部仿射结构．每个数据点用邻居点的线性组合来近似．于是通过寻找保持局部近似的最优方式构造低维表示．细节非常有趣，所以在这里给出：
 
-1. 对每个 $p$ 维中的数据点 $x_i$，寻找欧式距离的 $K$ 最近邻 $\cal N(i)$
+1. 对每个 $p$ 维中的数据点 $x_i$，寻找欧式距离意义下的 $K$ 最近邻点 $\cal N(i)$.
 2. 对每个点用邻居点的混合仿射来近似
 $$
-\underset{W_{ik}}\Vert x_i-\sum\limits_{k\in\cal N(i)}w_{ik}x_k\Vert^2\tag{14.102}
+\underset{w_{ik}}{\min}\Vert x_i-\sum\limits_{k\in\cal N(i)}w_{ik}x_k\Vert^2\tag{14.102}
 $$
 其中权重 $w_{ik}$ 满足 $w_{ik}=0, k\not\in \cal N(i), \sum_{k=1}^Nw_{ik}=1$．$w_{ik}$ 是点 $k$ 对 $i$ 点的重构的贡献．注意到为了得到唯一解，我们必须要求 $K < p$．
 3. 最后，固定 $w_{ik}$，在 $d < p$ 维空间中寻找点 $y_i$ 来最小化
@@ -36,9 +36,12 @@ $$
 \tr [(\Y-\W\Y)^T(\Y-\W\Y)] = \tr[\Y^T(\I-\W)^T(\I-\W)\Y)]\tag{14.104}
 $$
 
-其中 $\W$ 是 $N\times N$; $\Y$ 是 $N\times d, d < p$．$\hat \Y$ 的解是 $\M=(\I-\W)^T(\I-\W)$ 的 trailing eigenvectors([Issue 59](https://github.com/szcf-weiya/ESL-CN/issues/59))．因为 $\1$ 是特征值为 0 的平凡特征向量，所以我们舍弃它并且保留接下来的 $d$ 个．这会产生额外的影响 $\1^T\Y=0$，因此嵌入坐标(embedding coordinates)进行了中心化．
+其中 $\W$ 是 $N\times N$; $\Y$ 是 $N\times d, d < p$．$\hat \Y$ 的解是 $\M=(\I-\W)^T(\I-\W)$ 的 **尾特征向量 (trailing eigenvectors)**．因为 $\1$ 是特征值为 0 的平凡特征向量，所以我们舍弃它并且保留接下来的 $d$ 个．这会产生额外的影响 $\1^T\Y=0$，因此嵌入坐标的均值为0．
 
-**局部 MDS**（Chen and Buja, 2008[^3]） 采用最简单的、而且可以说是最直接的方式．定义 $\cal N$ 为邻居点的对称集；具体地，如果点 $i$ 在 $i'$ 的 $K$ 最近邻中，则点对 $(i, i')$ 在 $\cal N$ 中，反过来也是如此．
+!!! note "weiya 注:"
+	这里的尾特征向量除去了特征值为 0 的平凡特征向量 $\1$，而因为特征向量间正交，所以有 $\1^T\Y=0$. 
+
+**局部多维缩放 (Local MDS)**（Chen and Buja, 2008[^3]）采用最简单的、而且可以说是最直接的方式．定义 $\cal N$ 为邻居点的对称集；具体地，如果点 $i$ 在 $i'$ 的 $K$ 最近邻中，则点对 $(i, i')$ 在 $\cal N$ 中，反过来也是如此．
 
 !!! note "weiya 注"
 	在[14.7节](14.7-Independent-Component-Analysis-and-Exploratory-Projection-Pursuit/index.html)中的谱聚类的 mutual K-nearest-neighbor graph 也有用到 $\cal N$．
@@ -46,24 +49,27 @@ $$
 于是我们构造压力函数
 
 $$
-S_L(z_1,z_2,\ldots, z_N) = \sum\limits_{(i,i')\in \cal N}(d_{ii'}-\Vert z_i-z_{i'}\Vert)^2 + \sum\limits_{(i,i')\not\in \cal N}w\cdot (D-\Vert z_i-z_{i'}\Vert)^2\tag{14.105}
+S_L(z_1,z_2,\ldots, z_N) = \sum\limits_{(i,i')\in \cal N}(d_{ii'}-\Vert z_i-z_{i'}\Vert)^2 + \sum\limits_{(i,i')\not\in \cal N}w\cdot (D-\Vert z_i-z_{i'}\Vert)^2\tag{14.105}\label{14.105}
 $$
 
-这里$D$是某个较大的常数，$w$是权重．想法是将不是邻居的点看成是距离非常远；这些点对被赋予小权重$w$使得它们不会主导整个压力函数．为了简化表达式，取$w\sim 1/D$，并令$D\rightarrow \infty$．展开式(14.105)，得到
+这里 $D$ 是某个较大的常数，$w$ 是权重．想法是将不是邻居的点看成是距离非常远；这些点对被赋予小权重 $w$ 使得它们不会主导整个压力函数．为了简化表达式，取 $w\sim 1/D$，并令 $D\rightarrow \infty$．展开式 \eqref{14.105}，得到
 
 $$
-S_L(z_1,z_2,\ldots, z_N)=\sum\limits_{(i, i')\in\cal N}(d_{ii'}-\Vert z_i-z_{i'})^2-\tau \sum\limits_{(i,i')\not \in \cal N}\Vert z_i-z_{i'}\Vert\tag{14.106}
+S_L(z_1,z_2,\ldots, z_N)=\sum\limits_{(i, i')\in\cal N}(d_{ii'}-\Vert z_i-z_{i'}\Vert)^2-\tau \sum\limits_{(i,i')\not \in \cal N}\Vert z_i-z_{i'}\Vert\tag{14.106}\label{14.106}
 $$
 
-其中$\tau =2wD$．式(14.106)试图保持数据的局部性质，而第二项促使非邻居对$(i, i')$的$z_i,z_{i'}$更远．局部MDS在固定邻居个数$K$以及调整参数$\tau$的情况下，在$z_i$上最小化压力函数(14.106)．
+其中 $\tau =2wD$．式 \eqref{14.106} 试图保持数据的局部性质，而第二项促使非邻居对 $(i, i')$ 的 $z_i,z_{i'}$ 更远．局部多维缩放在固定邻居个数 $K$ 以及调整参数 $\tau$ 的情况下，在 $z_i$ 上最小化压力函数 \eqref{14.106}．
 
-图14.44的右图显示了采用$k=2$个邻居和$\tau = 0.01$的局部MDS的结果．我们采用多个起始值的坐标下降来寻找(非凸)损失函数一个好的最小值．沿着曲线的点的顺序大部分都被保持了．
+图 14.44 的右图显示了采用 $k=2$ 个邻居和 $\tau = 0.01$ 的局部多维缩放的结果．我们采用多个起始值的 **坐标下降 (coordinate descent)** 来寻找（非凸）损失函数一个好的最小值．沿着曲线的点的顺序大部分都被保持了．
 
 ![](../img/14/fig14.45.png)
 
-图14.45显示了LLE方法的一个有趣的应用．数据包含1965张图象，数字化为$20\times 28$的灰白图象．图中展示了LLE的前两个坐标结果，它们解释了摆放位置以及表情的一些变异．类似的图象可以通过局部MDS得到．
+图 14.45 显示了 LLE 方法的一个有趣的应用．数据包含 1965 张图象，数字化为 $20\times 28$ 的灰白图象．图中展示了 LLE 的前两个坐标结果，它们解释了摆放位置以及表情的一些变异．类似的图象可以通过局部多维缩放得到．
+
+!!! note "原书脚注: "
+	Sam Roweis 和 Lawrence Saul 友好地提供了图 14.45.
 
-在Chen and Buja(2008)[^3]报告的实验中，局部MDS与ISOMAP和LLE相比表现得更好．他们也演示了局部MDS在图象布局方面很有用的应用．有些方法与这里讨论的方法有着紧密的联系，如谱聚类(14.5.3节)和核PCA(14.5.4节)．
+在 Chen and Buja(2008)[^3] 报告的实验中，局部多维缩放与 ISOMAP 和 LLE 相比表现得更好．他们也演示了局部多维缩放在图象布局方面很有用的应用．有些方法与这里讨论的方法有着紧密的联系，如谱聚类（[14.5.3 节](14.5-Principal-Components-Curves-and-Surfaces/index.html)）和核主成分（[14.5.4 节](14.5-Principal-Components-Curves-and-Surfaces/index.html)）．
 
 [^1]: Tenenbaum, J. B., de Silva, V. and Langford, J. C. (2000). A global geometric framework for nonlinear dimensionality reduction, Science 290: 2319–2323.
 [^2]: Roweis, S. T. and Saul, L. K. (2000). Locally linear embedding, Science 290: 2323–2326.
diff --git a/docs/tag.md b/docs/tag.md
index 84f42b9c82..40d3f631b9 100644
--- a/docs/tag.md
+++ b/docs/tag.md
@@ -11,8 +11,8 @@
 - alternative maximization procedure: 轮换最大化过程 ([第 8.5 节](08-Model-Inference-and-Averaging/8.5-The-EM-Algorithm/index.html))
 - alternate: 轮换 ([第 8.5 节](08-Model-Inference-and-Averaging/8.5-The-EM-Algorithm/index.html))
 - automatic flexible: 自动灵活 ([第 9.1 节](09-Additive-Models-Trees-and-Related-Methods/9.1-Generalized-Additive-Models/index.html))
-- abundance: 多度 ([第 10.14 节](10-Boosting-and-Additive-Trees/10.14-Illustrations/index.html))
 - aggregated data: 聚合数据 ([第 10.14 节](10-Boosting-and-Additive-Trees/10.14-Illustrations/index.html))
+- abundance: 多度 ([第 10.14 节](10-Boosting-and-Additive-Trees/10.14-Illustrations/index.html))
 - additive expansions: 加性展开 ([第 10.2 节](10-Boosting-and-Additive-Trees/10.2-Boosting-Fits-an-Additive-Model/index.html))
 - adequacy: 充分性 ([第 14.1 节](14-Unsupervised-Learning/14.1-Introduction/index.html))
 - Association rule analysis: 关联规则分析 ([第 14.2 节](14-Unsupervised-Learning/14.2-Association-Rules/index.html))
@@ -25,14 +25,14 @@
 
 ## B
 - bias: 偏差 ([第 2.3 节](02-Overview-of-Supervised-Learning/2.3-Two-Simple-Approaches-to-Prediction/index.html))
-- Bayes classifier: 贝叶斯分类 ([第 2.4 节](02-Overview-of-Supervised-Learning/2.4-Statistical-Decision-Theory/index.html))
 - Bayes rate: 贝叶斯阶 ([第 2.4 节](02-Overview-of-Supervised-Learning/2.4-Statistical-Decision-Theory/index.html))
+- Bayes classifier: 贝叶斯分类 ([第 2.4 节](02-Overview-of-Supervised-Learning/2.4-Statistical-Decision-Theory/index.html))
 - basis set: 基础集 ([第 7.6 节](07-Model-Assessment-and-Selection/7.6-The-Effective-Number-of-Parameters/index.html))
 - Bayes factor: 贝叶斯因子 ([第 7.7 节](07-Model-Assessment-and-Selection/7.7-The-Bayesian-Approach-and-BIC/index.html))
 - BIC: 贝叶斯信息准则 ([第 7.7 节](07-Model-Assessment-and-Selection/7.7-The-Bayesian-Approach-and-BIC/index.html))
 - bootstrap: 自助法 ([第 8.7 节](08-Model-Inference-and-Averaging/8.7-Bagging/index.html))
-- Bootstrap aggregation: 自助法整合 ([第 8.7 节](08-Model-Inference-and-Averaging/8.7-Bagging/index.html))
 - bag: 打包 ([第 8.7 节](08-Model-Inference-and-Averaging/8.7-Bagging/index.html))
+- Bootstrap aggregation: 自助法整合 ([第 8.7 节](08-Model-Inference-and-Averaging/8.7-Bagging/index.html))
 - Bay Area: 海湾地区 ([第 10.14 节](10-Boosting-and-Additive-Trees/10.14-Illustrations/index.html))
 - black box: 黑箱方法 ([第 10.7 节](10-Boosting-and-Additive-Trees/10.7-Off-the-Shelf-Procedures-for-Data-Mining/index.html))
 - blurred: 模糊的 ([第 12.7 节](12-Support-Vector-Machines-and-Flexible-Discriminants/12.7-Mixture-Discriminant-Analysis/index.html))
@@ -42,63 +42,66 @@
 - Boltzmann machines: 玻尔兹曼机 ([第 17.4 节](17-Undirected-Graphical-Models/17.4-Undirected-Graphical-Models-for-Discrete-Variables/index.html))
 
 ## C
-- categories: 类别型 ([第 2.2 节](02-Overview-of-Supervised-Learning/2.2-Variable-Types-and-Terminology/index.html))
 - classification: 分类 ([第 2.2 节](02-Overview-of-Supervised-Learning/2.2-Variable-Types-and-Terminology/index.html), [第 13.1 节](13-Prototype-Methods-and-Nearest-Neighbors/13.1-Introduction/index.html))
+- categories: 类别型 ([第 2.2 节](02-Overview-of-Supervised-Learning/2.2-Variable-Types-and-Terminology/index.html))
 - curse of dimensionality: 维度的诅咒 ([第 2.5 节](02-Overview-of-Supervised-Learning/2.5-Local-Methods-in-High-Dimensions/index.html))
 - closed: 闭型 ([第 2.6 节](02-Overview-of-Supervised-Learning/2.6-Statistical-Models-Supervised-Learning-and-Function-Approximation/index.html))
 - cubic smoothing spline: 三次光滑样条 ([第 2.8 节](02-Overview-of-Supervised-Learning/2.8-Classes-of-Restricted-Estimators/index.html))
 - complexity: 复杂性 ([第 2.9 节](02-Overview-of-Supervised-Learning/2.9-Model-Selection-and-the-Bias-Variance-Tradeoff/index.html))
 - computer era: 计算机时代 ([第 3.1 节](03-Linear-Methods-for-Regression/3.1-Introduction/index.html))
 - closed form: 闭形式 ([第 3.4 节](03-Linear-Methods-for-Regression/3.4-Shrinkage-Methods/index.html))
+- canonical correlation analysis, CCA: 典则相关分析 ([第 3.7 节](03-Linear-Methods-for-Regression/3.7-Multiple-Outcome-Shrinkage-and-Selection/index.html))
 - compressed sensing: 压缩传感 ([第 3.8 节](03-Linear-Methods-for-Regression/3.8-More-on-the-Lasso-and-Related-Path-Algorithms/index.html))
-- canonical: 典则 ([第 4.3 节](04-Linear-Methods-for-Classification/4.3-Linear-Discriminant-Analysis/index.html))
 - canonical variables: 典则变量 ([第 4.3 节](04-Linear-Methods-for-Classification/4.3-Linear-Discriminant-Analysis/index.html))
-- cases: 案例集 ([第 4.4 节](04-Linear-Methods-for-Classification/4.4-Logistic-Regression/index.html))
-- controls: 控制集 ([第 4.4 节](04-Linear-Methods-for-Classification/4.4-Logistic-Regression/index.html))
+- canonical: 典则 ([第 4.3 节](04-Linear-Methods-for-Classification/4.3-Linear-Discriminant-Analysis/index.html))
 - concave: 凹的 ([第 4.4 节](04-Linear-Methods-for-Classification/4.4-Logistic-Regression/index.html))
+- controls: 控制集 ([第 4.4 节](04-Linear-Methods-for-Classification/4.4-Logistic-Regression/index.html))
+- cases: 案例集 ([第 4.4 节](04-Linear-Methods-for-Classification/4.4-Logistic-Regression/index.html))
 - compact kernel: 紧核 ([第 6.1 节](06-Kernel-Smoothing-Methods/6.1-One-Dimensional-Kernel-Smoothers/index.html))
-- consensus vote: 投票共识 ([第 8.7 节](08-Model-Inference-and-Averaging/8.7-Bagging/index.html))
 - consensus vote: 共识投票 ([第 8.7 节](08-Model-Inference-and-Averaging/8.7-Bagging/index.html))
+- consensus vote: 投票共识 ([第 8.7 节](08-Model-Inference-and-Averaging/8.7-Bagging/index.html))
 - collection of rules: 规则集合 ([第 9.3 节](09-Additive-Models-Trees-and-Related-Methods/9.3-PRIM/index.html))
 - computationally intensive: 计算密集型的 ([第 10.2 节](10-Boosting-and-Additive-Trees/10.2-Boosting-Fits-an-Additive-Model/index.html))
 - checkerboard: 跳棋盘 ([第 13.3 节](13-Prototype-Methods-and-Nearest-Neighbors/13.3-k-Nearest-Neighbor-Classifiers/index.html))
 - condensing: 压缩 ([第 13.5 节](13-Prototype-Methods-and-Nearest-Neighbors/13.5-Computational-Considerations/index.html))
-- conjunctive rule: 联合规则 ([第 14.2 节](14-Unsupervised-Learning/14.2-Association-Rules/index.html))
 - cardinality: 基数 ([第 14.2 节](14-Unsupervised-Learning/14.2-Association-Rules/index.html))
 - confidence: 置信度 ([第 14.2 节](14-Unsupervised-Learning/14.2-Association-Rules/index.html))
-- codebook: 码本 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
+- conjunctive rule: 联合规则 ([第 14.2 节](14-Unsupervised-Learning/14.2-Association-Rules/index.html))
+- Complete linkage, CL: 全链接 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
 - combinatorial algorithms: 组合算法 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
-- cluster analysis: 聚类分析 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
+- codebook: 码本 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
 - cophenetic correlation coefficient: 共表型相关系数 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
+- cluster analysis: 聚类分析 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
 - constrained topological map: 约束拓扑图 ([第 14.4 节](14-Unsupervised-Learning/14.4-Self-Organizing-Maps/index.html))
 - corpus callosum: 胼胝体 ([第 14.5 节](14-Unsupervised-Learning/14.5-Principal-Components-Curves-and-Surfaces/index.html))
 - convex hull: 凸包 ([第 14.6 节](14-Unsupervised-Learning/14.6-Non-negative-Matrix-Factorization/index.html))
 - correlated: 相关的 ([第 14.7 节](14-Unsupervised-Learning/14.7-Independent-Component-Analysis-and-Exploratory-Projection-Pursuit/index.html))
 - classical scaling: 经典缩放 ([第 14.8 节](14-Unsupervised-Learning/14.8-Multidimensional-Scaling/index.html))
+- coordinate descent: 坐标下降 ([第 14.9 节](14-Unsupervised-Learning/14.9-Nonlinear-Dimension-Reduction-and-Local-Multidimensional-Scaling/index.html))
 - clique: 团 ([第 17.2 节](17-Undirected-Graphical-Models/17.2-Markov-Graphs-and-Their-Properties/index.html))
 - clique potentials: 团势 ([第 17.2 节](17-Undirected-Graphical-Models/17.2-Markov-Graphs-and-Their-Properties/index.html))
 - complete graph: 完全图 ([第 17.2 节](17-Undirected-Graphical-Models/17.2-Markov-Graphs-and-Their-Properties/index.html))
 - covariance: 协方差图 ([第 17.3 节](17-Undirected-Graphical-Models/17.3-Undirected-Graphical-Models-for-Continuous-Variables/index.html))
 - constant: 常值 ([第 17.4 节](17-Undirected-Graphical-Models/17.4-Undirected-Graphical-Models-for-Discrete-Variables/index.html))
-- contrastive divergence: 对比发散 ([第 17.4 节](17-Undirected-Graphical-Models/17.4-Undirected-Graphical-Models-for-Discrete-Variables/index.html))
-- cyclical coordinate descent: 循环坐标下降 ([第 17.4 节](17-Undirected-Graphical-Models/17.4-Undirected-Graphical-Models-for-Discrete-Variables/index.html))
 - CD: 对比发散 ([第 17.4 节](17-Undirected-Graphical-Models/17.4-Undirected-Graphical-Models-for-Discrete-Variables/index.html))
+- cyclical coordinate descent: 循环坐标下降 ([第 17.4 节](17-Undirected-Graphical-Models/17.4-Undirected-Graphical-Models-for-Discrete-Variables/index.html))
+- contrastive divergence: 对比发散 ([第 17.4 节](17-Undirected-Graphical-Models/17.4-Undirected-Graphical-Models-for-Discrete-Variables/index.html))
 - cyclical coordinate descent: 坐标轮换 ([第 18.4 节](18-High-Dimensional-Problems/18.4-Linear-Classifiers-with-L1-Regularization/index.html))
 
 ## D
 - dependent variables: 因变量 ([第 2.1 节](02-Overview-of-Supervised-Learning/2.1-Introduction/index.html))
-- discrete: 离散 ([第 2.2 节](02-Overview-of-Supervised-Learning/2.2-Variable-Types-and-Terminology/index.html))
 - dummy variables: 虚拟变量 ([第 2.2 节](02-Overview-of-Supervised-Learning/2.2-Variable-Types-and-Terminology/index.html), [第 14.2 节](14-Unsupervised-Learning/14.2-Association-Rules/index.html))
+- discrete: 离散 ([第 2.2 节](02-Overview-of-Supervised-Learning/2.2-Variable-Types-and-Terminology/index.html))
 - dictionary: 字典 ([第 2.8 节](02-Overview-of-Supervised-Learning/2.8-Classes-of-Restricted-Estimators/index.html))
 - diamond: 菱形 ([第 3.4 节](03-Linear-Methods-for-Regression/3.4-Shrinkage-Methods/index.html))
-- discriminant functions: 判别函数 ([第 4.1 节](04-Linear-Methods-for-Classification/4.1-Introduction/index.html))
 - decision boundaries: 线性判别边界 ([第 4.1 节](04-Linear-Methods-for-Classification/4.1-Introduction/index.html))
+- discriminant functions: 判别函数 ([第 4.1 节](04-Linear-Methods-for-Classification/4.1-Introduction/index.html))
 - discriminant coordinates: 判别坐标 ([第 4.3 节](04-Linear-Methods-for-Classification/4.3-Linear-Discriminant-Analysis/index.html))
-- discriminant: 判别 ([第 4.3 节](04-Linear-Methods-for-Classification/4.3-Linear-Discriminant-Analysis/index.html))
 - discriminant variable: 判别变量 ([第 4.3 节](04-Linear-Methods-for-Classification/4.3-Linear-Discriminant-Analysis/index.html))
+- discriminant: 判别 ([第 4.3 节](04-Linear-Methods-for-Classification/4.3-Linear-Discriminant-Analysis/index.html))
 - driven: 驱动 ([第 5.8 节](05-Basis-Expansions-and-Regularization/5.8-Regularization-and-Reproducing-Kernel-Hibert-Spaces/index.html))
-- dictionaries of basis functions: 基函数的字典集 ([第 5.9 节](05-Basis-Expansions-and-Regularization/5.9-Wavelet-Smoothing/index.html))
 - dilation: 伸缩 ([第 5.9 节](05-Basis-Expansions-and-Regularization/5.9-Wavelet-Smoothing/index.html))
+- dictionaries of basis functions: 基函数的字典集 ([第 5.9 节](05-Basis-Expansions-and-Regularization/5.9-Wavelet-Smoothing/index.html))
 - divided differences: 差商 ([第 Appendix 节](05-Basis-Expansions-and-Regularization/Appendix-Computations-for-B-splines/index.html))
 - deviance: 偏差 ([第 7.2 节](07-Model-Assessment-and-Selection/7.2-Bias-Variance-and-Model-Complexity/index.html), [第 10.5 节](10-Boosting-and-Additive-Trees/10.5-Why-Exponential-Loss/index.html))
 - demographics: 人口统计数据 ([第 10.14 节](10-Boosting-and-Additive-Trees/10.14-Illustrations/index.html))
@@ -106,10 +109,10 @@
 - derived features: 导出特征 ([第 11.1 节](11-Neural-Networks/11.1-Introduction/index.html))
 - digitized analog signals: 数字化模拟信号 ([第 12.6 节](12-Support-Vector-Machines-and-Flexible-Discriminants/12.6-Penalized-Discriminant-Analysis/index.html))
 - DANN: 判别自适应最近邻 ([第 13.4 节](13-Prototype-Methods-and-Nearest-Neighbors/13.4-Adaptive-Nearest-Neighbor-Methods/index.html))
-- dissimilarities: 不相似性 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
+- dendrogram: 谱系图 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
 - decoding: 解码 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
+- dissimilarities: 不相似性 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
 - data segmentation: 数据分离 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
-- dendrogram: 谱系图 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
 - degree: 度 ([第 14.5 节](14-Unsupervised-Learning/14.5-Principal-Components-Curves-and-Surfaces/index.html))
 - differential entropy: 相对熵 ([第 14.7 节](14-Unsupervised-Learning/14.7-Independent-Component-Analysis-and-Exploratory-Projection-Pursuit/index.html))
 - dissimilarity: 不相似性 ([第 14.8 节](14-Unsupervised-Learning/14.8-Multidimensional-Scaling/index.html))
@@ -125,8 +128,8 @@
 - effective: 有效 ([第 2.3 节](02-Overview-of-Supervised-Learning/2.3-Two-Simple-Approaches-to-Prediction/index.html))
 - equivalent kernel: 等价核 ([第 2.7 节](02-Overview-of-Supervised-Learning/2.7-Structured-Regression-Models/index.html), [第 6.1 节](06-Kernel-Smoothing-Methods/6.1-One-Dimensional-Kernel-Smoothers/index.html))
 - effect size: 有效大小 ([第 3.2 节](03-Linear-Methods-for-Regression/3.2-Linear-Regression-Models-and-Least-Squares/index.html))
-- effective degrees of freedom: 有效自由度 ([第 3.4 节](03-Linear-Methods-for-Regression/3.4-Shrinkage-Methods/index.html), [第 5.4 节](05-Basis-Expansions-and-Regularization/5.4-Smoothing-Splines/index.html))
 - eigen decomposition: 特征值分解 ([第 3.4 节](03-Linear-Methods-for-Regression/3.4-Shrinkage-Methods/index.html))
+- effective degrees of freedom: 有效自由度 ([第 3.4 节](03-Linear-Methods-for-Regression/3.4-Shrinkage-Methods/index.html), [第 5.4 节](05-Basis-Expansions-and-Regularization/5.4-Smoothing-Splines/index.html))
 - equivalent kernels: 等价核 ([第 5.4 节](05-Basis-Expansions-and-Regularization/5.4-Smoothing-Splines/index.html))
 - EPE: 积分平方预测误差 ([第 5.5 节](05-Basis-Expansions-and-Regularization/5.5-Automatic-Selection-of-the-Smoothing-Parameters/index.html))
 - evaluation: 赋值 ([第 6.0 节](06-Kernel-Smoothing-Methods/6.0-Introduction/index.html))
@@ -138,8 +141,8 @@
 - experts: 专家 ([第 9.5 节](09-Additive-Models-Trees-and-Related-Methods/9.5-Hierarchical-Mixtures-of-Experts/index.html))
 - early stopping: 早停 ([第 10.12 节](10-Boosting-and-Additive-Trees/10.12-Regularization/index.html))
 - empirical risk: 经验风险 ([第 10.9 节](10-Boosting-and-Additive-Trees/10.9-Boosting-Trees/index.html))
-- editing: 编辑 ([第 13.5 节](13-Prototype-Methods-and-Nearest-Neighbors/13.5-Computational-Considerations/index.html))
 - exterior point: 外点 ([第 13.5 节](13-Prototype-Methods-and-Nearest-Neighbors/13.5-Computational-Considerations/index.html))
+- editing: 编辑 ([第 13.5 节](13-Prototype-Methods-and-Nearest-Neighbors/13.5-Computational-Considerations/index.html))
 - effectiveness: 有效性 ([第 14.1 节](14-Unsupervised-Learning/14.1-Introduction/index.html))
 - encoding: 编码 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
 - exploratory projection pursuit: 探索投影寻踪 ([第 14.7 节](14-Unsupervised-Learning/14.7-Independent-Component-Analysis-and-Exploratory-Projection-Pursuit/index.html))
@@ -159,8 +162,8 @@
 - forward stagewise: 向前逐步 ([第 10.9 节](10-Boosting-and-Additive-Trees/10.9-Boosting-Trees/index.html))
 - flexible discriminant analysis: 可变的判别分析 ([第 12.1 节](12-Support-Vector-Machines-and-Flexible-Discriminants/12.1-Introduction/index.html))
 - feature vector: 特征向量 ([第 14.2 节](14-Unsupervised-Learning/14.2-Association-Rules/index.html))
-- factor loadings: 因子载荷 ([第 14.7 节](14-Unsupervised-Learning/14.7-Independent-Component-Analysis-and-Exploratory-Projection-Pursuit/index.html))
 - Factor Analysis: 因子分析 ([第 14.7 节](14-Unsupervised-Learning/14.7-Independent-Component-Analysis-and-Exploratory-Projection-Pursuit/index.html))
+- factor loadings: 因子载荷 ([第 14.7 节](14-Unsupervised-Learning/14.7-Independent-Component-Analysis-and-Exploratory-Projection-Pursuit/index.html))
 - feature selection: 特征选择 ([第 18.1 节](18-High-Dimensional-Problems/18.1-When-p-is-Much-Bigger-than-N/index.html))
 
 ## G
@@ -171,10 +174,11 @@
 - generalized linear models: 广义线性模型 ([第 6.5 节](06-Kernel-Smoothing-Methods/6.5-Local-Likelihood-and-Other-Models/index.html))
 - generalization error: 泛化误差 ([第 7.2 节](07-Model-Assessment-and-Selection/7.2-Bias-Variance-and-Model-Complexity/index.html))
 - gating networks: 门控网络 ([第 9.5 节](09-Additive-Models-Trees-and-Related-Methods/9.5-Hierarchical-Mixtures-of-Experts/index.html))
+- Group Average, GA: 群平均 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
 - generalized additive spline model: 广义可加样条模型 ([第 14.7 节](14-Unsupervised-Learning/14.7-Independent-Component-Analysis-and-Exploratory-Projection-Pursuit/index.html))
 - Graph: 图 ([第 17.1 节](17-Undirected-Graphical-Models/17.1-Introduction/index.html))
-- global Markov properties: 全局马尔科夫性质 ([第 17.2 节](17-Undirected-Graphical-Models/17.2-Markov-Graphs-and-Their-Properties/index.html))
 - global Markov properties: 整体马尔科夫性质 ([第 17.2 节](17-Undirected-Graphical-Models/17.2-Markov-Graphs-and-Their-Properties/index.html))
+- global Markov properties: 全局马尔科夫性质 ([第 17.2 节](17-Undirected-Graphical-Models/17.2-Markov-Graphs-and-Their-Properties/index.html))
 - Gradient descent: 梯度下降 ([第 17.4 节](17-Undirected-Graphical-Models/17.4-Undirected-Graphical-Models-for-Discrete-Variables/index.html))
 - gene expression arrays: 基因表达阵列 ([第 18.2 节](18-High-Dimensional-Problems/18.2-Diagonal-Linear-Discriminant-Analysis-and-Nearest-Shrunken-Centroids/index.html))
 
@@ -186,9 +190,9 @@
 - having no say: 不起作用 ([第 6.7 节](06-Kernel-Smoothing-Methods/6.7-Radial-Basis-Functions-and-Kernels/index.html))
 - holes: 洞 ([第 6.7 节](06-Kernel-Smoothing-Methods/6.7-Radial-Basis-Functions-and-Kernels/index.html))
 - HME: 混合层次专家 ([第 9.0 节](09-Additive-Models-Trees-and-Related-Methods/9.0-Introduction/index.html))
+- hard decision: 硬决定 ([第 9.5 节](09-Additive-Models-Trees-and-Related-Methods/9.5-Hierarchical-Mixtures-of-Experts/index.html))
 - hierarchical mixtures of experts: 专家的分层混合 ([第 9.5 节](09-Additive-Models-Trees-and-Related-Methods/9.5-Hierarchical-Mixtures-of-Experts/index.html))
 - HME: 专家的分层混合 ([第 9.5 节](09-Additive-Models-Trees-and-Related-Methods/9.5-Hierarchical-Mixtures-of-Experts/index.html))
-- hard decision: 硬决定 ([第 9.5 节](09-Additive-Models-Trees-and-Related-Methods/9.5-Hierarchical-Mixtures-of-Experts/index.html))
 - highly skewed: 高偏的 ([第 10.7 节](10-Boosting-and-Additive-Trees/10.7-Off-the-Shelf-Procedures-for-Data-Mining/index.html))
 - hidden units: 隐藏层 ([第 11.3 节](11-Neural-Networks/11.3-Neural-Networks/index.html))
 - hierarchical basis: 分层基 ([第 12.3 节](12-Support-Vector-Machines-and-Flexible-Discriminants/12.3-Support-Vector-Machines-and-Kernels/index.html))
@@ -217,23 +221,24 @@
 - invariance manifolds: 不变流形 ([第 13.3 节](13-Prototype-Methods-and-Nearest-Neighbors/13.3-k-Nearest-Neighbor-Classifiers/index.html))
 - invariant metric: 不变度量 ([第 13.3 节](13-Prototype-Methods-and-Nearest-Neighbors/13.3-k-Nearest-Neighbor-Classifiers/index.html))
 - items set: 项目集 ([第 14.2 节](14-Unsupervised-Learning/14.2-Association-Rules/index.html))
-- ICA: 独立成分分析 ([第 14.7 节](14-Unsupervised-Learning/14.7-Independent-Component-Analysis-and-Exploratory-Projection-Pursuit/index.html))
 - Independent Component Analysis: 独立成分分析 ([第 14.7 节](14-Unsupervised-Learning/14.7-Independent-Component-Analysis-and-Exploratory-Projection-Pursuit/index.html))
+- ICA: 独立成分分析 ([第 14.7 节](14-Unsupervised-Learning/14.7-Independent-Component-Analysis-and-Exploratory-Projection-Pursuit/index.html))
 - isotonic regression: 保序回归 ([第 14.8 节](14-Unsupervised-Learning/14.8-Multidimensional-Scaling/index.html))
-- ISOMAP: 等距特征映射算法 ([第 14.9 节](14-Unsupervised-Learning/14.9-Nonlinear-Dimension-Reduction-and-Local-Multidimensional-Scaling/index.html))
+- Isometric feature mapping, ISOMAP: 等距特征映射算法 ([第 14.9 节](14-Unsupervised-Learning/14.9-Nonlinear-Dimension-Reduction-and-Local-Multidimensional-Scaling/index.html))
 - importance sampling: 重要性采样 ([第 16.3 节](16-Ensemble-Learning/16.3-Learning-Ensembles/index.html))
 - Importance Sampling: 重要度采样 ([第 16.3 节](16-Ensemble-Learning/16.3-Learning-Ensembles/index.html))
 - inference: 推断 ([第 17.1 节](17-Undirected-Graphical-Models/17.1-Introduction/index.html))
 - iterative proportional fitting procedure: 迭代比例拟合过程 ([第 17.3 节](17-Undirected-Graphical-Models/17.3-Undirected-Graphical-Models-for-Continuous-Variables/index.html))
 - iteratively reweighted least squares: 迭代重赋权最小二乘法 ([第 17.4 节](17-Undirected-Graphical-Models/17.4-Undirected-Graphical-Models-for-Discrete-Variables/index.html))
+- Iterative proportional fitting, IPF: 迭代比例过滤 ([第 17.4 节](17-Undirected-Graphical-Models/17.4-Undirected-Graphical-Models-for-Discrete-Variables/index.html))
 - independent rule: 独立规则 ([第 18.2 节](18-High-Dimensional-Problems/18.2-Diagonal-Linear-Discriminant-Analysis-and-Nearest-Shrunken-Centroids/index.html))
 
 ## J
 - joint maximization: 联合最大化 ([第 8.5 节](08-Model-Inference-and-Averaging/8.5-The-EM-Algorithm/index.html))
 
 ## K
-- kernel function: 核函数 ([第 2.8 节](02-Overview-of-Supervised-Learning/2.8-Classes-of-Restricted-Estimators/index.html))
 - knots: 结点 ([第 2.8 节](02-Overview-of-Supervised-Learning/2.8-Classes-of-Restricted-Estimators/index.html))
+- kernel function: 核函数 ([第 2.8 节](02-Overview-of-Supervised-Learning/2.8-Classes-of-Restricted-Estimators/index.html))
 - kernel property: 核性质 ([第 5.8 节](05-Basis-Expansions-and-Regularization/5.8-Regularization-and-Reproducing-Kernel-Hibert-Spaces/index.html))
 - kernel reproducing property: 核再生性质 ([第 5.8 节](05-Basis-Expansions-and-Regularization/5.8-Regularization-and-Reproducing-Kernel-Hibert-Spaces/index.html))
 - kernel: 核 ([第 6.0 节](06-Kernel-Smoothing-Methods/6.0-Introduction/index.html))
@@ -242,21 +247,21 @@
 ## L
 - large: 大 ([第 2.2 节](02-Overview-of-Supervised-Learning/2.2-Variable-Types-and-Terminology/index.html))
 - loss function: 损失函数 ([第 2.4 节](02-Overview-of-Supervised-Learning/2.4-Statistical-Decision-Theory/index.html))
-- learning by example: 样本学习 ([第 2.6 节](02-Overview-of-Supervised-Learning/2.6-Statistical-Models-Supervised-Learning-and-Function-Approximation/index.html))
 - linear basis expansions: 线性基展开 ([第 2.6 节](02-Overview-of-Supervised-Learning/2.6-Statistical-Models-Supervised-Learning-and-Function-Approximation/index.html))
+- learning by example: 样本学习 ([第 2.6 节](02-Overview-of-Supervised-Learning/2.6-Statistical-Models-Supervised-Learning-and-Function-Approximation/index.html))
 - learning problem: 学习问题 ([第 2 章文献笔记](02-Overview-of-Supervised-Learning/Bibliographic-Notes/index.html))
 - least squares: 最小二乘 ([第 3.2 节](03-Linear-Methods-for-Regression/3.2-Linear-Regression-Models-and-Least-Squares/index.html))
-- LAR: 最小角回归 ([第 3.4 节](03-Linear-Methods-for-Regression/3.4-Shrinkage-Methods/index.html))
 - Lagrangian form: 拉格朗日形式 ([第 3.4 节](03-Linear-Methods-for-Regression/3.4-Shrinkage-Methods/index.html))
+- LAR: 最小角回归 ([第 3.4 节](03-Linear-Methods-for-Regression/3.4-Shrinkage-Methods/index.html))
 - LDA: 线性判别分析 ([第 4.3 节](04-Linear-Methods-for-Classification/4.3-Linear-Discriminant-Analysis/index.html))
 - linear basis expansion: 线性基展开式 ([第 5.1 节](05-Basis-Expansions-and-Regularization/5.1-Introduction/index.html))
 - linear smoother: 线性光滑 ([第 5.4 节](05-Basis-Expansions-and-Regularization/5.4-Smoothing-Splines/index.html))
 - localized in time and in frequency: 在时间和在频率上局部化 ([第 5.9 节](05-Basis-Expansions-and-Regularization/5.9-Wavelet-Smoothing/index.html))
 - low scale: 尺度最低 ([第 5.9 节](05-Basis-Expansions-and-Regularization/5.9-Wavelet-Smoothing/index.html))
-- lag set: 滞后集 ([第 6.5 节](06-Kernel-Smoothing-Methods/6.5-Local-Likelihood-and-Other-Models/index.html))
-- local linear logistic model: 局部线性逻辑斯蒂回归模型 ([第 6.5 节](06-Kernel-Smoothing-Methods/6.5-Local-Likelihood-and-Other-Models/index.html))
 - logistic: 逻辑斯蒂回归 ([第 6.5 节](06-Kernel-Smoothing-Methods/6.5-Local-Likelihood-and-Other-Models/index.html))
 - local regression: 局部回归 ([第 6.5 节](06-Kernel-Smoothing-Methods/6.5-Local-Likelihood-and-Other-Models/index.html))
+- lag set: 滞后集 ([第 6.5 节](06-Kernel-Smoothing-Methods/6.5-Local-Likelihood-and-Other-Models/index.html))
+- local linear logistic model: 局部线性逻辑斯蒂回归模型 ([第 6.5 节](06-Kernel-Smoothing-Methods/6.5-Local-Likelihood-and-Other-Models/index.html))
 - light fitting: 轻拟合 ([第 7.11 节](07-Model-Assessment-and-Selection/7.11-Bootstrap-Methods/index.html))
 - LOO: 舍一法 ([第 7.12 节](07-Model-Assessment-and-Selection/7.12-Conditional-or-Expected-Test-Error/index.html))
 - latent class model: 潜类别模型 ([第 9.5 节](09-Additive-Models-Trees-and-Related-Methods/9.5-Hierarchical-Mixtures-of-Experts/index.html))
@@ -266,7 +271,8 @@
 - left singular vectors: 左奇异向量 ([第 14.5 节](14-Unsupervised-Learning/14.5-Principal-Components-Curves-and-Surfaces/index.html))
 - loadings: 因子载荷 ([第 14.7 节](14-Unsupervised-Learning/14.7-Independent-Component-Analysis-and-Exploratory-Projection-Pursuit/index.html))
 - least squares scaling: 最小二乘缩放 ([第 14.8 节](14-Unsupervised-Learning/14.8-Multidimensional-Scaling/index.html))
-- LLE: 局部线性内嵌 ([第 14.9 节](14-Unsupervised-Learning/14.9-Nonlinear-Dimension-Reduction-and-Local-Multidimensional-Scaling/index.html))
+- Local MDS: 局部多维缩放 ([第 14.9 节](14-Unsupervised-Learning/14.9-Nonlinear-Dimension-Reduction-and-Local-Multidimensional-Scaling/index.html))
+- Local linear embedding, LLE: 局部线性内嵌 ([第 14.9 节](14-Unsupervised-Learning/14.9-Nonlinear-Dimension-Reduction-and-Local-Multidimensional-Scaling/index.html))
 - least angle regression: 最小角回归 ([第 16.2 节](16-Ensemble-Learning/16.2-Boosting-and-Regularization-Paths/index.html))
 - learning: 学习 ([第 17.1 节](17-Undirected-Graphical-Models/17.1-Introduction/index.html))
 - less fitting is better: 欠拟合更好 ([第 18.1 节](18-High-Dimensional-Problems/18.1-When-p-is-Much-Bigger-than-N/index.html))
@@ -278,8 +284,8 @@
 - model complexity: 模型复杂度 ([第 2.9 节](02-Overview-of-Supervised-Learning/2.9-Model-Selection-and-the-Bias-Variance-Tradeoff/index.html))
 - mean squared error: 均方误差 ([第 2.9 节](02-Overview-of-Supervised-Learning/2.9-Model-Selection-and-the-Bias-Variance-Tradeoff/index.html))
 - model selection: 模型选择 ([第 3.3 节](03-Linear-Methods-for-Regression/3.3-Subset-Selection/index.html))
-- marginal likelihood: 边缘似然 ([第 4.4 节](04-Linear-Methods-for-Classification/4.4-Logistic-Regression/index.html))
 - multinomial: 多项式分布 ([第 4.4 节](04-Linear-Methods-for-Classification/4.4-Logistic-Regression/index.html))
+- marginal likelihood: 边缘似然 ([第 4.4 节](04-Linear-Methods-for-Classification/4.4-Logistic-Regression/index.html))
 - margin: 边缘 ([第 4.5 节](04-Linear-Methods-for-Classification/4.5-Separating-Hyperplanes/index.html))
 - margin: 空白 ([第 4.5 节](04-Linear-Methods-for-Classification/4.5-Separating-Hyperplanes/index.html))
 - moving beyond linearity: 超线性 ([第 5.1 节](05-Basis-Expansions-and-Regularization/5.1-Introduction/index.html))
@@ -290,8 +296,8 @@
 - Markov chain Monte Carlo: 马尔科夫蒙特卡洛法 ([第 8.6 节](08-Model-Inference-and-Averaging/8.6-MCMC-for-Sampling-from-the-Posterior/index.html))
 - mode: 最大值 ([第 8.8 节](08-Model-Inference-and-Averaging/8.8-Model-Averaging-and-Stacking/index.html))
 - MARS: 多元自适应回归样条 ([第 9.0 节](09-Additive-Models-Trees-and-Related-Methods/9.0-Introduction/index.html))
-- MAR: 随机缺失 ([第 9.6 节](09-Additive-Models-Trees-and-Related-Methods/9.6-Missing-Data/index.html))
 - MCAR: 完全随机缺失 ([第 9.6 节](09-Additive-Models-Trees-and-Related-Methods/9.6-Missing-Data/index.html))
+- MAR: 随机缺失 ([第 9.6 节](09-Additive-Models-Trees-and-Related-Methods/9.6-Missing-Data/index.html))
 - multiple additive regression trees: 多重可加回归树 ([第 10.10 节](10-Boosting-and-Additive-Trees/10.10-Numerical-Optimization-via-Gradient-Boosting/index.html))
 - marginal average: 边缘平均 ([第 10.13 节](10-Boosting-and-Additive-Trees/10.13-Interpretation/index.html))
 - median house values: 房子价值的中位数 ([第 10.14 节](10-Boosting-and-Additive-Trees/10.14-Illustrations/index.html))
@@ -303,17 +309,18 @@
 - mixing proportions: 混合比例 ([第 12.7 节](12-Support-Vector-Machines-and-Flexible-Discriminants/12.7-Mixture-Discriminant-Analysis/index.html))
 - manifold: 流形 ([第 14.1 节](14-Unsupervised-Learning/14.1-Introduction/index.html))
 - mode finding: 模式寻找 ([第 14.2 节](14-Unsupervised-Learning/14.2-Association-Rules/index.html))
-- mixture modeling: 混合模型 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
 - melanomas: 黑素瘤 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
 - mode seeking: 模式寻找 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
+- mixture modeling: 混合模型 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
 - missing values: 缺失值 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
 - mutual information: 互信息量 ([第 14.7 节](14-Unsupervised-Learning/14.7-Independent-Component-Analysis-and-Exploratory-Projection-Pursuit/index.html))
 - metric: 度量 ([第 14.8 节](14-Unsupervised-Learning/14.8-Multidimensional-Scaling/index.html))
+- MDS, Multidimensional scaling: 多维缩放 ([第 14.8 节](14-Unsupervised-Learning/14.8-Multidimensional-Scaling/index.html))
 - Markov random fields: 马尔科夫随机域 ([第 17.1 节](17-Undirected-Graphical-Models/17.1-Introduction/index.html))
 - Markov networks: 马尔科夫网络 ([第 17.1 节](17-Undirected-Graphical-Models/17.1-Introduction/index.html))
 - multiple testing: 多重检验 ([第 18.1 节](18-High-Dimensional-Problems/18.1-When-p-is-Much-Bigger-than-N/index.html), [第 18.7 节](18-High-Dimensional-Problems/18.7-Feature-Assessment-and-the-Multiple-Testing-Problem/index.html))
-- maximal margin classifier: 最大边界分类器 ([第 18.3 节](18-High-Dimensional-Problems/18.3-Linear-Classifiers-with-Quadratic-Regularization/index.html))
 - margin tree: 边际树 ([第 18.3 节](18-High-Dimensional-Problems/18.3-Linear-Classifiers-with-Quadratic-Regularization/index.html))
+- maximal margin classifier: 最大边界分类器 ([第 18.3 节](18-High-Dimensional-Problems/18.3-Linear-Classifiers-with-Quadratic-Regularization/index.html))
 - maximal margin solution: 最大边缘解 ([第 18.5 节](18-High-Dimensional-Problems/18.5-Classification-When-Features-are-Unavailable/index.html))
 
 ## N
@@ -329,10 +336,11 @@
 - nonlinear approximating manifolds: 线性逼近流形 ([第 14.5 节](14-Unsupervised-Learning/14.5-Principal-Components-Curves-and-Surfaces/index.html))
 - negentropy: 负熵 ([第 14.7 节](14-Unsupervised-Learning/14.7-Independent-Component-Analysis-and-Exploratory-Projection-Pursuit/index.html))
 - numerical quadrature: 数值积分 ([第 16.3 节](16-Ensemble-Learning/16.3-Learning-Ensembles/index.html))
-- nodes: 结点 ([第 17.4 节](17-Undirected-Graphical-Models/17.4-Undirected-Graphical-Models-for-Discrete-Variables/index.html))
 - Newton updates: 牛顿更新 ([第 17.4 节](17-Undirected-Graphical-Models/17.4-Undirected-Graphical-Models-for-Discrete-Variables/index.html))
+- nodes: 结点 ([第 17.4 节](17-Undirected-Graphical-Models/17.4-Undirected-Graphical-Models-for-Discrete-Variables/index.html))
 - nearest centroid classifer: 最近重心分类器 ([第 18.2 节](18-High-Dimensional-Problems/18.2-Diagonal-Linear-Discriminant-Analysis-and-Nearest-Shrunken-Centroids/index.html))
 - nearest centroid classifier: 最近重心分类器 ([第 18.2 节](18-High-Dimensional-Problems/18.2-Diagonal-Linear-Discriminant-Analysis-and-Nearest-Shrunken-Centroids/index.html))
+- nearest shrunken centroids, NSC: 最近收缩重心 ([第 18.2 节](18-High-Dimensional-Problems/18.2-Diagonal-Linear-Discriminant-Analysis-and-Nearest-Shrunken-Centroids/index.html))
 
 ## O
 - outputs: 输出变量 ([第 2.1 节](02-Overview-of-Supervised-Learning/2.1-Introduction/index.html))
@@ -355,8 +363,8 @@
 - perceptron: 感知器 ([第 4.1 节](04-Linear-Methods-for-Classification/4.1-Introduction/index.html))
 - pair of classes: 类别对 ([第 4.3 节](04-Linear-Methods-for-Classification/4.3-Linear-Discriminant-Analysis/index.html))
 - parsimonious: 最简洁 ([第 4.4 节](04-Linear-Methods-for-Classification/4.4-Logistic-Regression/index.html))
-- perceptron learning algorithm: 感知器学习算法 ([第 4.5 节](04-Linear-Methods-for-Classification/4.5-Separating-Hyperplanes/index.html))
 - perceptrons: 感知器 ([第 4.5 节](04-Linear-Methods-for-Classification/4.5-Separating-Hyperplanes/index.html))
+- perceptron learning algorithm: 感知器学习算法 ([第 4.5 节](04-Linear-Methods-for-Classification/4.5-Separating-Hyperplanes/index.html))
 - pairwise variance: 逐点方差 ([第 5.2 节](05-Basis-Expansions-and-Regularization/5.2-Piecewise-Polynomials-and-Splines/index.html))
 - penalty matrix: 惩罚矩阵 ([第 5.4 节](05-Basis-Expansions-and-Regularization/5.4-Smoothing-Splines/index.html))
 - projection: 投影 ([第 5.4 节](05-Basis-Expansions-and-Regularization/5.4-Smoothing-Splines/index.html))
@@ -370,9 +378,9 @@
 - peeled: 剔除 ([第 9.3 节](09-Additive-Models-Trees-and-Related-Methods/9.3-PRIM/index.html))
 - pseudo residuals: 伪残差 ([第 10.10 节](10-Boosting-and-Additive-Trees/10.10-Numerical-Optimization-via-Gradient-Boosting/index.html))
 - primitive: 原型 ([第 10.11 节](10-Boosting-and-Additive-Trees/10.11-Right-Sized-Trees-for-Boosting/index.html))
-- partial dependence: 偏相依性 ([第 10.13 节](10-Boosting-and-Additive-Trees/10.13-Interpretation/index.html))
-- purely additive: 纯可加的 ([第 10.13 节](10-Boosting-and-Additive-Trees/10.13-Interpretation/index.html))
 - partial: 偏 ([第 10.13 节](10-Boosting-and-Additive-Trees/10.13-Interpretation/index.html))
+- purely additive: 纯可加的 ([第 10.13 节](10-Boosting-and-Additive-Trees/10.13-Interpretation/index.html))
+- partial dependence: 偏相依性 ([第 10.13 节](10-Boosting-and-Additive-Trees/10.13-Interpretation/index.html))
 - purely multiplicative: 纯可乘的 ([第 10.13 节](10-Boosting-and-Additive-Trees/10.13-Interpretation/index.html))
 - Pacific coast: 太平洋沿岸 ([第 10.14 节](10-Boosting-and-Additive-Trees/10.14-Illustrations/index.html))
 - predictive learning: 预测学习 ([第 10.7 节](10-Boosting-and-Additive-Trees/10.7-Off-the-Shelf-Procedures-for-Data-Mining/index.html))
@@ -388,16 +396,16 @@
 - principal curves and surfaces: 主曲线和主曲面 ([第 14.4 节](14-Unsupervised-Learning/14.4-Self-Organizing-Maps/index.html))
 - Principal points: 主点 ([第 14.5 节](14-Unsupervised-Learning/14.5-Principal-Components-Curves-and-Surfaces/index.html))
 - projection matrix: 投影矩阵 ([第 14.5 节](14-Unsupervised-Learning/14.5-Principal-Components-Curves-and-Surfaces/index.html))
-- Percentage Squared Prediction Explained: 解释的方差的百分比 ([第 16.2 节](16-Ensemble-Learning/16.2-Boosting-and-Regularization-Paths/index.html))
 - Percentage Misclassification Error Explained: 解释的误分类率的百分比 ([第 16.2 节](16-Ensemble-Learning/16.2-Boosting-and-Regularization-Paths/index.html))
+- Percentage Squared Prediction Explained: 解释的方差的百分比 ([第 16.2 节](16-Ensemble-Learning/16.2-Boosting-and-Regularization-Paths/index.html))
 - potential: 势 ([第 17.1 节](17-Undirected-Graphical-Models/17.1-Introduction/index.html))
-- pairwise Markov independencies: 逐对马尔科夫独立 ([第 17.2 节](17-Undirected-Graphical-Models/17.2-Markov-Graphs-and-Their-Properties/index.html))
 - partition: 分割 ([第 17.2 节](17-Undirected-Graphical-Models/17.2-Markov-Graphs-and-Their-Properties/index.html))
-- potential function: 势函数 ([第 17.2 节](17-Undirected-Graphical-Models/17.2-Markov-Graphs-and-Their-Properties/index.html))
 - partition function: 分割函数 ([第 17.2 节](17-Undirected-Graphical-Models/17.2-Markov-Graphs-and-Their-Properties/index.html), [第 17.4 节](17-Undirected-Graphical-Models/17.4-Undirected-Graphical-Models-for-Discrete-Variables/index.html))
 - path: 路径 ([第 17.2 节](17-Undirected-Graphical-Models/17.2-Markov-Graphs-and-Their-Properties/index.html))
-- pairwise Markov graphs: 成对马尔科夫图 ([第 17.2 节](17-Undirected-Graphical-Models/17.2-Markov-Graphs-and-Their-Properties/index.html))
 - pairwise Markov properties: 逐对马尔科夫性质 ([第 17.2 节](17-Undirected-Graphical-Models/17.2-Markov-Graphs-and-Their-Properties/index.html))
+- potential function: 势函数 ([第 17.2 节](17-Undirected-Graphical-Models/17.2-Markov-Graphs-and-Their-Properties/index.html))
+- pairwise Markov graphs: 成对马尔科夫图 ([第 17.2 节](17-Undirected-Graphical-Models/17.2-Markov-Graphs-and-Their-Properties/index.html))
+- pairwise Markov independencies: 逐对马尔科夫独立 ([第 17.2 节](17-Undirected-Graphical-Models/17.2-Markov-Graphs-and-Their-Properties/index.html))
 - partial covariances: 偏协方差 ([第 17.3 节](17-Undirected-Graphical-Models/17.3-Undirected-Graphical-Models-for-Continuous-Variables/index.html))
 - prior probability: 先验概率 ([第 18.2 节](18-High-Dimensional-Problems/18.2-Diagonal-Linear-Discriminant-Analysis-and-Nearest-Shrunken-Centroids/index.html))
 - posterior mode: 后验的众数 ([第 18.4 节](18-High-Dimensional-Problems/18.4-Linear-Classifiers-with-L1-Regularization/index.html))
@@ -426,8 +434,8 @@
 - regression splines: 回归样条 ([第 5.2 节](05-Basis-Expansions-and-Regularization/5.2-Piecewise-Polynomials-and-Splines/index.html))
 - radial basis functions: 径向基函数 ([第 5.7 节](05-Basis-Expansions-and-Regularization/5.7-Multidimensional-Splines/index.html))
 - reproducing kernel Hilbert space: 再生核希尔伯特空间 ([第 5.8 节](05-Basis-Expansions-and-Regularization/5.8-Regularization-and-Reproducing-Kernel-Hibert-Spaces/index.html))
-- RKHS: 再生核希尔伯特空间 ([第 5.8 节](05-Basis-Expansions-and-Regularization/5.8-Regularization-and-Reproducing-Kernel-Hibert-Spaces/index.html))
 - reproducing kernel Hilbert spaces: 再生核希尔伯特空间 ([第 5.8 节](05-Basis-Expansions-and-Regularization/5.8-Regularization-and-Reproducing-Kernel-Hibert-Spaces/index.html), [第 12.3 节](12-Support-Vector-Machines-and-Flexible-Discriminants/12.3-Support-Vector-Machines-and-Kernels/index.html))
+- RKHS: 再生核希尔伯特空间 ([第 5.8 节](05-Basis-Expansions-and-Regularization/5.8-Regularization-and-Reproducing-Kernel-Hibert-Spaces/index.html))
 - reference: 参考 ([第 5.9 节](05-Basis-Expansions-and-Regularization/5.9-Wavelet-Smoothing/index.html))
 - relative overfitting rate: 相对过拟合率 ([第 7.11 节](07-Model-Assessment-and-Selection/7.11-Bootstrap-Methods/index.html))
 - responsibilities: 责任 ([第 8.6 节](08-Model-Inference-and-Averaging/8.6-MCMC-for-Sampling-from-the-Posterior/index.html))
@@ -436,9 +444,9 @@
 - ridge function: 岭函数 ([第 11.2 节](11-Neural-Networks/11.2-Projection-Pursuit-Regression/index.html))
 - radical basis function network: 径向基函数网络 ([第 11.3 节](11-Neural-Networks/11.3-Neural-Networks/index.html))
 - regions: 区域 ([第 14.2 节](14-Unsupervised-Learning/14.2-Association-Rules/index.html))
-- rotation: 旋转 ([第 14.5 节](14-Unsupervised-Learning/14.5-Principal-Components-Curves-and-Surfaces/index.html))
 - reconstruction error: 重构误差 ([第 14.5 节](14-Unsupervised-Learning/14.5-Principal-Components-Curves-and-Surfaces/index.html))
 - right singular vectors: 右奇异向量 ([第 14.5 节](14-Unsupervised-Learning/14.5-Principal-Components-Curves-and-Surfaces/index.html))
+- rotation: 旋转 ([第 14.5 节](14-Unsupervised-Learning/14.5-Principal-Components-Curves-and-Surfaces/index.html))
 - relevance network: 相关网络 ([第 17.3 节](17-Undirected-Graphical-Models/17.3-Undirected-Graphical-Models-for-Continuous-Variables/index.html))
 - RBM: 限制玻尔兹曼机 ([第 17.4 节](17-Undirected-Graphical-Models/17.4-Undirected-Graphical-Models-for-Discrete-Variables/index.html))
 - RDA: 正则化判别分析 ([第 18.3 节](18-High-Dimensional-Problems/18.3-Linear-Classifiers-with-Quadratic-Regularization/index.html))
@@ -449,44 +457,47 @@
 ## S
 - supervised learning: 监督学习 ([第 2.1 节](02-Overview-of-Supervised-Learning/2.1-Introduction/index.html), [第 14.1 节](14-Unsupervised-Learning/14.1-Introduction/index.html))
 - small: 小 ([第 2.2 节](02-Overview-of-Supervised-Learning/2.2-Variable-Types-and-Terminology/index.html))
-- squared error loss: 平方误差损失 ([第 2.4 节](02-Overview-of-Supervised-Learning/2.4-Statistical-Decision-Theory/index.html))
 - simultaneously: 同时 ([第 2.4 节](02-Overview-of-Supervised-Learning/2.4-Statistical-Decision-Theory/index.html))
+- squared error loss: 平方误差损失 ([第 2.4 节](02-Overview-of-Supervised-Learning/2.4-Statistical-Decision-Theory/index.html))
 - smoothing: 光滑化 ([第 2.8 节](02-Overview-of-Supervised-Learning/2.8-Classes-of-Restricted-Estimators/index.html))
 - smoothing: 光滑 ([第 2.9 节](02-Overview-of-Supervised-Learning/2.9-Model-Selection-and-the-Bias-Variance-Tradeoff/index.html))
-- stationary condition: 平稳条件 ([第 3.4 节](03-Linear-Methods-for-Regression/3.4-Shrinkage-Methods/index.html))
-- subset selection: 子集选择 ([第 3.4 节](03-Linear-Methods-for-Regression/3.4-Shrinkage-Methods/index.html))
 - SVD: 奇异值分解 ([第 3.4 节](03-Linear-Methods-for-Regression/3.4-Shrinkage-Methods/index.html))
 - shrinkage methods: 收缩方法 ([第 3.4 节](03-Linear-Methods-for-Regression/3.4-Shrinkage-Methods/index.html))
+- subset selection: 子集选择 ([第 3.4 节](03-Linear-Methods-for-Regression/3.4-Shrinkage-Methods/index.html))
+- stationary condition: 平稳条件 ([第 3.4 节](03-Linear-Methods-for-Regression/3.4-Shrinkage-Methods/index.html))
 - scale invariant: 尺度不变 ([第 3.5 节](03-Linear-Methods-for-Regression/3.5-Methods-Using-Derived-Input-Directions/index.html))
 - simple coordinate descent: 简单坐标下降 ([第 3.8 节](03-Linear-Methods-for-Regression/3.8-More-on-the-Lasso-and-Related-Path-Algorithms/index.html))
+- smoothly clipped absolute deviation, SCAD: 平稳削减绝对偏差法 ([第 3.8 节](03-Linear-Methods-for-Regression/3.8-More-on-the-Lasso-and-Related-Path-Algorithms/index.html))
+- stochastic gradient descent: 随机梯度下降 ([第 4.5 节](04-Linear-Methods-for-Classification/4.5-Separating-Hyperplanes/index.html))
 - separating hyperplane classifiers: 分离超平面分类器 ([第 4.5 节](04-Linear-Methods-for-Classification/4.5-Separating-Hyperplanes/index.html))
-- support vector machine: 支持向量机 ([第 4.5 节](04-Linear-Methods-for-Classification/4.5-Separating-Hyperplanes/index.html), [第 12.1 节](12-Support-Vector-Machines-and-Flexible-Discriminants/12.1-Introduction/index.html), [第 12.3 节](12-Support-Vector-Machines-and-Flexible-Discriminants/12.3-Support-Vector-Machines-and-Kernels/index.html))
 - slab: 平板 ([第 4.5 节](04-Linear-Methods-for-Classification/4.5-Separating-Hyperplanes/index.html))
-- stochastic gradient descent: 随机梯度下降 ([第 4.5 节](04-Linear-Methods-for-Classification/4.5-Separating-Hyperplanes/index.html))
 - support points: 支撑点 ([第 4.5 节](04-Linear-Methods-for-Classification/4.5-Separating-Hyperplanes/index.html))
+- support vector machine: 支持向量机 ([第 4.5 节](04-Linear-Methods-for-Classification/4.5-Separating-Hyperplanes/index.html), [第 12.1 节](12-Support-Vector-Machines-and-Flexible-Discriminants/12.1-Introduction/index.html), [第 12.3 节](12-Support-Vector-Machines-and-Flexible-Discriminants/12.3-Support-Vector-Machines-and-Kernels/index.html))
 - splines: 样条 ([第 5.1 节](05-Basis-Expansions-and-Regularization/5.1-Introduction/index.html))
+- smoothing parameter: 光滑参数 ([第 5.4 节](05-Basis-Expansions-and-Regularization/5.4-Smoothing-Splines/index.html))
 - smoother matrix: 光滑矩阵 ([第 5.4 节](05-Basis-Expansions-and-Regularization/5.4-Smoothing-Splines/index.html))
 - shrinking: 收缩 ([第 5.4 节](05-Basis-Expansions-and-Regularization/5.4-Smoothing-Splines/index.html))
-- smoothing parameter: 光滑参数 ([第 5.4 节](05-Basis-Expansions-and-Regularization/5.4-Smoothing-Splines/index.html))
 - spurious: 假的 ([第 5.7 节](05-Basis-Expansions-and-Regularization/5.7-Multidimensional-Splines/index.html))
 - sparse: 稀疏 ([第 5.9 节](05-Basis-Expansions-and-Regularization/5.9-Wavelet-Smoothing/index.html), [第 16.2 节](16-Ensemble-Learning/16.2-Boosting-and-Regularization-Paths/index.html))
 - span: 跨度 ([第 6.2 节](06-Kernel-Smoothing-Methods/6.2-Selecting-the-Width-of-the-Kernel/index.html))
-- sample standard deviation: 样本标准偏差 ([第 7.10 节](07-Model-Assessment-and-Selection/7.10-Cross-Validation/index.html))
 - standard error: 标准误差 ([第 7.10 节](07-Model-Assessment-and-Selection/7.10-Cross-Validation/index.html))
+- sample standard deviation: 样本标准偏差 ([第 7.10 节](07-Model-Assessment-and-Selection/7.10-Cross-Validation/index.html))
+- SRM, structural risk minimization: 结构风险最小化 ([第 7.9 节](07-Model-Assessment-and-Selection/7.9-Vapnik-Chervonenkis-Dimension/index.html))
 - score function: 得分函数 ([第 8.2 节](08-Model-Inference-and-Averaging/8.2-The-Bootstrap-and-Maximum-Likelihood-Methods/index.html))
 - Stacked generalization: 堆栈泛化 ([第 8.8 节](08-Model-Inference-and-Averaging/8.8-Model-Averaging-and-Stacking/index.html))
 - soft splits: 软分割 ([第 9.5 节](09-Additive-Models-Trees-and-Related-Methods/9.5-Hierarchical-Mixtures-of-Experts/index.html))
 - soft probabilistic: 软概率的决定 ([第 9.5 节](09-Additive-Models-Trees-and-Related-Methods/9.5-Hierarchical-Mixtures-of-Experts/index.html))
 - surrogate splits: 代理分割 ([第 9.6 节](09-Additive-Models-Trees-and-Related-Methods/9.6-Missing-Data/index.html))
-- steepest descent: 最速下降 ([第 10.10 节](10-Boosting-and-Additive-Trees/10.10-Numerical-Optimization-via-Gradient-Boosting/index.html))
 - step length: 步长 ([第 10.10 节](10-Boosting-and-Additive-Trees/10.10-Numerical-Optimization-via-Gradient-Boosting/index.html))
+- steepest descent: 最速下降 ([第 10.10 节](10-Boosting-and-Additive-Trees/10.10-Numerical-Optimization-via-Gradient-Boosting/index.html))
 - squared relative importance: 平方相对重要度 ([第 10.13 节](10-Boosting-and-Additive-Trees/10.13-Interpretation/index.html))
 - salinity: 盐度 ([第 10.14 节](10-Boosting-and-Additive-Trees/10.14-Illustrations/index.html))
 - scaled classification tree: 缩放分类树 ([第 10.9 节](10-Boosting-and-Additive-Trees/10.9-Boosting-Trees/index.html))
 - single index model: 单指标模型 ([第 11.2 节](11-Neural-Networks/11.2-Projection-Pursuit-Regression/index.html))
 - stochastic approximation: 随机近似 ([第 11.4 节](11-Neural-Networks/11.4-Fitting-Neural-Networks/index.html))
-- statistics folklore: 统计民俗学 ([第 14.2 节](14-Unsupervised-Learning/14.2-Association-Rules/index.html))
 - support: 支撑集 ([第 14.2 节](14-Unsupervised-Learning/14.2-Association-Rules/index.html))
+- statistics folklore: 统计民俗学 ([第 14.2 节](14-Unsupervised-Learning/14.2-Association-Rules/index.html))
+- Single linkage, SL: 单链接 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
 - similarities: 相似性 ([第 14.3 节](14-Unsupervised-Learning/14.3-Cluster-Analysis/index.html))
 - stress function: 压力函数 ([第 14.8 节](14-Unsupervised-Learning/14.8-Multidimensional-Scaling/index.html))
 - separate: 分离 ([第 17.2 节](17-Undirected-Graphical-Models/17.2-Markov-Graphs-and-Their-Properties/index.html))
@@ -507,10 +518,10 @@
 - tensor product basis: 张量积基底 ([第 5.7 节](05-Basis-Expansions-and-Regularization/5.7-Multidimensional-Splines/index.html))
 - translation: 平移 ([第 5.9 节](05-Basis-Expansions-and-Regularization/5.9-Wavelet-Smoothing/index.html), [第 14.5 节](14-Unsupervised-Learning/14.5-Principal-Components-Curves-and-Surfaces/index.html))
 - time and frequency localization: 时间和频率的局部化 ([第 5.9 节](05-Basis-Expansions-and-Regularization/5.9-Wavelet-Smoothing/index.html))
-- tie: 结 ([第 6.1 节](06-Kernel-Smoothing-Methods/6.1-One-Dimensional-Kernel-Smoothers/index.html))
 - trimming the hills: 截断山坡 ([第 6.1 节](06-Kernel-Smoothing-Methods/6.1-One-Dimensional-Kernel-Smoothers/index.html))
-- Training error: 训练误差 ([第 7.2 节](07-Model-Assessment-and-Selection/7.2-Bias-Variance-and-Model-Complexity/index.html))
+- tie: 结 ([第 6.1 节](06-Kernel-Smoothing-Methods/6.1-One-Dimensional-Kernel-Smoothers/index.html))
 - tunning parameter: 调整参数 ([第 7.2 节](07-Model-Assessment-and-Selection/7.2-Bias-Variance-and-Model-Complexity/index.html))
+- Training error: 训练误差 ([第 7.2 节](07-Model-Assessment-and-Selection/7.2-Bias-Variance-and-Model-Complexity/index.html))
 - test error: 测试误差 ([第 7.2 节](07-Model-Assessment-and-Selection/7.2-Bias-Variance-and-Model-Complexity/index.html))
 - trees: 树 ([第 9.0 节](09-Additive-Models-Trees-and-Related-Methods/9.0-Introduction/index.html))
 - target function: 目标函数 ([第 10.11 节](10-Boosting-and-Additive-Trees/10.11-Right-Sized-Trees-for-Boosting/index.html))
@@ -518,6 +529,7 @@
 - the structured space of functions: 函数的结构空间 ([第 12.3 节](12-Support-Vector-Machines-and-Flexible-Discriminants/12.3-Support-Vector-Machines-and-Kernels/index.html))
 - tangent distance: 切线距离 ([第 13.3 节](13-Prototype-Methods-and-Nearest-Neighbors/13.3-k-Nearest-Neighbor-Classifiers/index.html))
 - tangent distance: 切向距离 ([第 13.5 节](13-Prototype-Methods-and-Nearest-Neighbors/13.5-Computational-Considerations/index.html))
+- trailing eigenvectors: 尾特征向量 ([第 14.9 节](14-Unsupervised-Learning/14.9-Nonlinear-Dimension-Reduction-and-Local-Multidimensional-Scaling/index.html))
 - time of flight: 时间长度 ([第 18.4 节](18-High-Dimensional-Problems/18.4-Linear-Classifiers-with-L1-Regularization/index.html))
 
 ## U
diff --git a/gentag.py b/gentag.py
index bbff0a0fc8..959e31f9f5 100644
--- a/gentag.py
+++ b/gentag.py
@@ -3,7 +3,7 @@
 import glob
 # ([\u4e00-\u9fa5]+): chinese translation
 # (\b[a-zA-Z ]+\b): original
-pat = re.compile(r"\*\*([\u4e00-\u9fa5]+)\s?\((\b[a-zA-Z ]+\b)\)\*\*")
+pat = re.compile(r"\*\*([\u4e00-\u9fa5]+)\s?\((\b[a-zA-Z ,]+\b)\)\*\*")
 
 tags = [[] for i in range(26)]
 docsdir = os.listdir("docs/")