Skip to content

M
mtbookv2
Commit fa380e17 by 孟霞
Options

4和9的图片名字

parent 3de58171
正在显示 包含 64 行增加 和 200 行删除
Chapter4/Figures/a-deeper-and-wider-grammar-tree-means-more-complex-sentence-structures.tex deleted 100644 → 0
\begin{tikzpicture}[scale=0.5]
\Tree[.IP
[.ADVP
[.AD \node(e1){前几天};]
]
[.LCP
[.NP
[.CP
[.IP
[.VP
[.VV \node(e2){出};]
[.NP
[.NN \node(e3){事故};]
]
]
]
[.DEC \node(e4){的};]
]
[.DP
[.DT \node(e5){那};]
[.CLP
[.M \node(e6){条};]
]
]
[.NP
[.NN \node(e7){街};]
]
]
[.LC \node(e8){上};]
]
[.VP
[.VP
[.VE \node(e9){有};]
[.IP
[.NP
[.NN \node(e10){一家};]
]
[.VP
[.ADVP
[.AD \node(e11){非常};]
]
[.VP
[.VV \node(e12){气派的};]
[.NP
[.ADJP
[.JJ \node(e13){大};]
]
[.NP
[.NN \node(e14){商店};]
]
]
]
]
]
]
[.VP
[.ADVP
[.AD \node(e15){，那里};]
]
[.ADVP
[.AD \node(e16){经常};]
]
[.VP
[.VV \node(e17){出售};]
[.NP
[.QP
[.CD \node(e18){一些};]
]
[.ADJP
[.JJ \node(e19){名贵};]
]
[.NP
[.NN \node(e20){鲜花};]
]
]
]
]
]
[.PU \node(e21){.};]
]
\end{tikzpicture}
\ No newline at end of file
Chapter4/Figures/a-shallow-and-narrow-grammar-means-a-simpler-sentence-structure.tex deleted 100644 → 0
\begin{tikzpicture}[scale=0.7]
\Tree[.IP
[.NP
[.NR \node(e1){俄国};]
]
[.VP
[.VV \node(e2){希望};]
[.IP
[.NP
[.NR \node(e3){伊朗};]
]
[.VP
[.ADVP
[.AD \node(e4) {没有};]
]
[.VP
[.VV \node(e5) {制造};]
[.NP
[.NN \node(e6){核武器};]
[.NN \node(e7){计划};]
]
]
]
]
]
[.PU \node(e8){.};]
]
\end{tikzpicture}
\ No newline at end of file
Chapter4/Figures/several-common-distributed-representations.tex deleted 100644 → 0
\begin{tikzpicture}
\tikzstyle{unit1} = [inner sep=1pt,align=center,font=\footnotesize]
\tikzstyle{unit2} = [inner sep=1pt,align=center,font=\scriptsize]
\node[unit1] (n1) at (0,0){单词分布式表示};
\node[unit2,anchor=west] (n11) at ([xshift=1em,yshift=4em]n1.east){one-hot词向量};
\node[unit2,anchor=west] (n12) at ([xshift=1em,yshift=2.4em]n1.east){Word2Vec词向量};
\node[unit2,anchor=west] (n13) at ([xshift=1em,yshift=0.8em]n1.east){GloVe词向量};
\node[unit2,anchor=west] (n14) at ([xshift=1em,yshift=-0.8em]n1.east){…};
\node[unit2,anchor=west] (n15) at ([xshift=1em,yshift=-2.4em]n1.east){ELMO预训练词向量};
\node[unit2,anchor=west] (n16) at ([xshift=1em,yshift=-4em]n1.east){BERT预训练词向量};
\draw[decorate,decoration={brace,mirror,amplitude=2mm}] ([xshift=-0.3em]n11.west) -- ([xshift=-0.3em]n16.west);
\node[unit1,anchor=west] (n2) at ([xshift=9em]n1.east){句子分布式表示};
\node[unit2,anchor=west] (n21) at ([xshift=1em,yshift=4.2em]n2.east){RAE编码};
\node[unit2,anchor=west] (n22) at ([xshift=1em,yshift=2.8em]n2.east){Doc2Vec向量};
\node[unit2,anchor=west] (n23) at ([xshift=1em,yshift=1.4em]n2.east){ELMO预训练句子表示};
\node[unit2,anchor=west] (n24) at ([xshift=1em,yshift=0em]n2.east){…};
\node[unit2,anchor=west] (n25) at ([xshift=1em,yshift=-1.4em]n2.east){GPT句子表示};
\node[unit2,anchor=west] (n26) at ([xshift=1em,yshift=-2.8em]n2.east){BERT预训练句子表示};
\node[unit2,anchor=west] (n27) at ([xshift=1em,yshift=-4.2em]n2.east){skip-thought向量};
\draw[decorate,decoration={brace,mirror,amplitude=2mm}] ([xshift=-0.3em]n21.west) -- ([xshift=-0.3em]n27.west);
\end{tikzpicture}
Chapter4/chapter4.tex
......@@ -88,7 +88,7 @@
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter4/Figures/logic-diagram-of-translation-quality-evaluation-method}
\input{./Chapter4/Figures/figure-logic-diagram-of-translation-quality-evaluation-method}
\caption{译文质量评价方法逻辑图}
\label{fig:4-2}
\end{figure}
......@@ -300,8 +300,8 @@
%----------------------------------------------
\begin{figure}[htp]
\centering
\subfigure[“绝对”匹配词对齐-1]{\input{./Chapter4/Figures/absolute-match-word-alignment-1}}
\subfigure[“绝对”匹配词对齐-2]{\input{./Chapter4/Figures/absolute-match-word-alignment-2}}
\subfigure[“绝对”匹配词对齐-1]{\input{./Chapter4/Figures/figure-absolute-match-word-alignment-1}}
\subfigure[“绝对”匹配词对齐-2]{\input{./Chapter4/Figures/figure-absolute-match-word-alignment-2}}
\caption{“绝对”匹配模型}
\label{fig:4-3}
\end{figure}
......@@ -313,7 +313,7 @@
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter4/Figures/match-words-with-stem}
\input{./Chapter4/Figures/figure-match-words-with-stem}
\caption{“波特词干”匹配词对齐}
\label{fig:4-4}
\end{figure}
......@@ -325,7 +325,7 @@
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter4/Figures/synonym-matching-word-alignment}
\input{./Chapter4/Figures/figure-synonym-matching-word-alignment}
\caption{“同义词”匹配词对齐}
\label{fig:4-5}
\end{figure}
......@@ -339,7 +339,7 @@
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter4/Figures/determine-final-word-alignment}
\input{./Chapter4/Figures/figure-determine-final-word-alignment}
\caption{确定最终词对齐}
\label{fig:4-6}
\end{figure}
......@@ -481,7 +481,7 @@ His house is on the south bank of the river.
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter4/Figures/representation-of-reference-answer-set-in-hyter}
\input{./Chapter4/Figures/figure-representation-of-reference-answer-set-in-hyter}
\caption{HyTER中参考答案集的表示方式}
\label{fig:4-7}
\end{figure}
......@@ -497,8 +497,8 @@ His house is on the south bank of the river.
%----------------------------------------------
\begin{figure}[htp]
\centering
\subfigure[英语参考答案集表示]{\input{./Chapter4/Figures/representation-of-english-reference-answer-set}}
\subfigure[捷克语参考答案集表示]{\input{./Chapter4/Figures/representation-of-czech-reference-answer-set}}
\subfigure[英语参考答案集表示]{\input{./Chapter4/Figures/figure-representation-of-english-reference-answer-set}}
\subfigure[捷克语参考答案集表示]{\input{./Chapter4/Figures/figure-representation-of-czech-reference-answer-set}}
\caption{使用HyTER构造的参考答案集}
\label{fig:4-8}
\end{figure}
......@@ -647,7 +647,7 @@ His house is on the south bank of the river.
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter4/Figures/The-process-of-statistical-hypothesis-testing}
\input{./Chapter4/Figures/figure-the-process-of-statistical-hypothesis-testing}
\caption{统计假设检验的流程}
\label{fig:4-13}
\end{figure}
......@@ -700,7 +700,7 @@ d=t \frac{s}{\sqrt{n}}
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter4/Figures/schematic-diagram-of-word-level-quality-assessment-task}
\input{./Chapter4/Figures/figure-schematic-diagram-of-word-level-quality-assessment-task}
\caption{单词级质量评估任务示意图}
\label{fig:4-11}
\end{figure}
......@@ -745,7 +745,7 @@ d=t \frac{s}{\sqrt{n}}
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter4/Figures/schematic-diagram-of-phrase-level-quality-assessment-task}
\input{./Chapter4/Figures/figure-schematic-diagram-of-phrase-level-quality-assessment-task}
\caption{短语级质量评估任务示意图}
\label{fig:4-12}
\end{figure}
......
Chapter9/chapter9.tex
......@@ -107,7 +107,7 @@
%----------------------------------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-the-amount-of-data-in-a-bilingual-corpus}
\input{./Chapter9/Figures/figure-the-amount-of-data-in-a-bilingual-corpus}
\caption{机器翻译系统所使用的双语数据量变化趋势}
\label{fig:9-1}
\end{figure}
......@@ -136,14 +136,14 @@
\subfigure[基于特征工程的机器学习方法做图像分类]{
\begin{minipage}{.9\textwidth}
\centering
\includegraphics[width=8cm]{./Chapter9/Figures/feature-engineering.jpg}
\includegraphics[width=8cm]{./Chapter9/Figures/figure-feature-engineering.jpg}
\end{minipage}%
}
\vfill
\subfigure[端到端学习方法做图像分类]{
\begin{minipage}{.9\textwidth}
\centering
\includegraphics[width=8cm]{./Chapter9/Figures/deep-learning.jpg}
\includegraphics[width=8cm]{./Chapter9/Figures/figure-deep-learning.jpg}
\end{minipage}%
}
\caption{特征工程{\small\sffamily\bfseries{vs}}端到端学习}
......@@ -513,7 +513,7 @@ l_p({\vectorn{\emph{x}}}) & = & {\Vert{\vectorn{\emph{x}}}\Vert}_p \nonumber \\
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-artificial-neuron}
\input{./Chapter9/Figures/figure-artificial-neuron}
\caption{人工神经元}
\label{fig:9-4}
\end{figure}
......@@ -536,7 +536,7 @@ y=\begin{cases} 0 & \sum_{i}{x_i\cdot w_i}-\sigma <0\\1 & \sum_{i}{x_i\cdot w_i}
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-perceptron-mode}
\input{./Chapter9/Figures/figure-perceptron-mode}
\caption{感知机模型}
\label{fig:9-5}
\end{figure}
......@@ -566,7 +566,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-perceptron-to-predict-1}
\input{./Chapter9/Figures/figure-perceptron-to-predict-1}
\caption{预测是否去剧场的感知机（权重相同）}
\label{fig:9-6}
\end{figure}
......@@ -590,7 +590,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-perceptron-to-predict-2}
\input{./Chapter9/Figures/figure-perceptron-to-predict-2}
\caption{预测是否去剧场的感知机（改变权重）}
\label{fig:9-7}
\end{figure}
......@@ -617,7 +617,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-different-forms-of-neuronal-input}
\input{./Chapter9/Figures/figure-different-forms-of-neuronal-input}
\caption{神经元输入的不同形式}
\label{fig:9-8}
\end{figure}
......@@ -644,7 +644,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-perceptron-to-predict-3}
\input{./Chapter9/Figures/figure-perceptron-to-predict-3}
\caption{预测是否去剧场的决策模型（只考虑女友喜好）}
\label{fig:9-9}
\end{figure}
......@@ -695,7 +695,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-corresponence-between-matrix-element-and-output}
\input{./Chapter9/Figures/figure-corresponence-between-matrix-element-and-output}
\caption{权重矩阵中的元素与输出的对应关系}
\label{fig:9-10}
\end{figure}
......@@ -706,7 +706,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-single-layer-of-neural-network-for-weather-prediction}
\input{./Chapter9/Figures/figure-single-layer-of-neural-network-for-weather-prediction}
\caption{预测天气的单层神经网络}
\label{fig:9-11}
\end{figure}
......@@ -734,7 +734,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-translation}
\input{./Chapter9/Figures/figure-translation}
\caption{线性变换示意图}
\label{fig:9-13}
\end{figure}
......@@ -746,7 +746,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-linear-transformation}
\input{./Chapter9/Figures/figure-linear-transformation}
\caption{线性变换3维$ \rightarrow $2维数学示意}
\label{fig:9-14}
\end{figure}
......@@ -758,7 +758,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-activate}
\input{./Chapter9/Figures/figure-activate}
\caption{几种常见的激活函数}
\label{fig:9-15}
\end{figure}
......@@ -776,7 +776,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-four-layers-of-neural-network}
\input{./Chapter9/Figures/figure-four-layers-of-neural-network}
\caption{具有四层神经元的（三层）神经网络}
\label{fig:9-17}
\end{figure}
......@@ -801,7 +801,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-two-layer-neural-network}
\input{./Chapter9/Figures/figure-two-layer-neural-network}
\caption{以Sigmoid作为隐藏层激活函数的两层神经网络}
\label{fig:9-18}
\end{figure}
......@@ -812,7 +812,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-weight}
\input{./Chapter9/Figures/figure-weight}
\caption{通过调整权重$ w_{11} $改变目标函数平滑程度}
\label{fig:9-19}
\end {figure}
......@@ -824,7 +824,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-bias}
\input{./Chapter9/Figures/figure-bias}
\caption{通过调整偏置量$ b_1 $改变目标函数位置}
\label{fig:9-20}
\end {figure}
......@@ -835,8 +835,8 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-w1}
\caption{通过改变权重$ w_{21} $将目标函数“拉高”或“压扁”}
\input{./Chapter9/Figures/figure-w1}
\caption{通过改变权重$ w'_{11} $将目标函数“拉高”或“压扁”}
\label{fig:9-21}
\end {figure}
%-------------------------------------------
......@@ -846,8 +846,8 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-w2}
\caption{通过设置第二组参数（$b_2$和$w_{22}$）将目标函数分段数增加}
\input{./Chapter9/Figures/figure-w2}
\caption{通过设置第二组参数（$b_2$和$w'_{21}$）将目标函数分段数增加}
\label{fig:9-22}
\end {figure}
%-------------------------------------------
......@@ -857,7 +857,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-piecewise}
\input{./Chapter9/Figures/figure-piecewise}
\caption{将目标函数作分段处理}
\label{fig:9-23}
\end {figure}
......@@ -870,7 +870,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-fit}
\input{./Chapter9/Figures/figure-fit}
\caption{扩展隐层神经元个数去拟合目标函数更多的“一小段”}
\label{fig:9-24}
\end {figure}
......@@ -927,7 +927,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-tensor-sample}
\input{./Chapter9/Figures/figure-tensor-sample}
\caption{3阶张量示例（$4 \times 4 \times 4$）}
\label{fig:9-25}
\end{figure}
......@@ -967,7 +967,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-tensor-mul}
\input{./Chapter9/Figures/figure-tensor-mul}
\caption{张量与矩阵的矩阵乘法}
\label{fig:9-27}
\end {figure}
......@@ -989,7 +989,7 @@ x_1\cdot w_1+x_2\cdot w_2+x_3\cdot w_3 & = & 0\cdot 1+0\cdot 1+1\cdot 1 \nonumbe
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-broadcast}
\input{./Chapter9/Figures/figure-broadcast}
\caption{广播机制}
\label{fig:9-28}
\end {figure}
......@@ -1026,7 +1026,7 @@ f(x)=\begin{cases} 0 & x\le 0 \\x & x>0\end{cases}
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-save}
\input{./Chapter9/Figures/figure-save}
\caption{1阶(a)、2阶(b)、3阶张量(c)的物理存储}
\label{fig:9-29}
\end{figure}
......@@ -1087,7 +1087,7 @@ f(x)=\begin{cases} 0 & x\le 0 \\x & x>0\end{cases}
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-weather}
\input{./Chapter9/Figures/figure-weather}
\caption{判断穿衣指数问题的神经网络过程}
\label{fig:9-37}
\end{figure}
......@@ -1103,7 +1103,7 @@ y&=&{\textrm{Sigmoid}}({\textrm{Tanh}}({\vectorn{\emph{x}}}\cdot {\vectorn{\emph
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-weather-forward}
\input{./Chapter9/Figures/figure-weather-forward}
\caption{前向计算示例（计算图）}
\label{fig:9-38}
\end{figure}
......@@ -1146,7 +1146,7 @@ y&=&{\textrm{Sigmoid}}({\textrm{Tanh}}({\vectorn{\emph{x}}}\cdot {\vectorn{\emph
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-absolute-loss}
\input{./Chapter9/Figures/figure-absolute-loss}
\caption{正确答案与神经网络输出之间的偏差}
\label{fig:9-42}
\end{figure}
......@@ -1207,7 +1207,7 @@ y&=&{\textrm{Sigmoid}}({\textrm{Tanh}}({\vectorn{\emph{x}}}\cdot {\vectorn{\emph
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-gradient-descent}
\input{./Chapter9/Figures/figure-gradient-descent}
\caption{函数上一个点沿着不同方向移动的示例}
\label{fig:9-43}
\end{figure}
......@@ -1367,7 +1367,7 @@ $+2x^2+x+1)$ & \ \ $(x^4+2x^3+2x^2+x+1)$ & $+6x+1$ \\
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-forward-propagation}
\input{./Chapter9/Figures/figure-forward-propagation}
\caption{前向计算示意图}
\label{fig:9-44}
\end{figure}
......@@ -1388,7 +1388,7 @@ $+2x^2+x+1)$ & \ \ $(x^4+2x^3+2x^2+x+1)$ & $+6x+1$ \\
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-back-propagation}
\input{./Chapter9/Figures/figure-back-propagation}
\caption{反向计算示意图}
\label{fig:9-45}
\end{figure}
......@@ -1435,7 +1435,7 @@ v_t&=&\beta v_{t-1}+(1-\beta)\frac{\partial J}{\partial \theta_t}
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-sawtooth}
\input{./Chapter9/Figures/figure-sawtooth}
\caption{Momentum梯度下降 vs 普通梯度下降}
\label{fig:9-46}
\end{figure}
......@@ -1529,7 +1529,7 @@ z_t&=&\gamma z_{t-1}+(1-\gamma) \frac{\partial J}{\partial {\theta}_t} \cdot \f
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-parallel}
\input{./Chapter9/Figures/figure-parallel}
\caption{同步更新与异步更新对比}
\label{fig:9-47}
\end {figure}
......@@ -1592,7 +1592,7 @@ z_t&=&\gamma z_{t-1}+(1-\gamma) \frac{\partial J}{\partial {\theta}_t} \cdot \f
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-residual-structure}
\input{./Chapter9/Figures/figure-residual-structure}
\caption{残差网络的结构}
\label{fig:9-51}
\end{figure}
......@@ -1637,7 +1637,7 @@ z_t&=&\gamma z_{t-1}+(1-\gamma) \frac{\partial J}{\partial {\theta}_t} \cdot \f
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-multilayer-neural-network-example}
\input{./Chapter9/Figures/figure-multilayer-neural-network-example}
\caption{多层神经网络实例}
\label{fig:9-52}
\end{figure}
......@@ -1699,7 +1699,7 @@ z_t&=&\gamma z_{t-1}+(1-\gamma) \frac{\partial J}{\partial {\theta}_t} \cdot \f
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-forward-propagation-output}
\input{./Chapter9/Figures/figure-forward-propagation-output}
\caption{输出层的前向计算过程}
\label{fig:9-53}
\end{figure}
......@@ -1723,7 +1723,7 @@ z_t&=&\gamma z_{t-1}+(1-\gamma) \frac{\partial J}{\partial {\theta}_t} \cdot \f
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-back-propagation-output1}
\input{./Chapter9/Figures/figure-back-propagation-output1}
\caption{从损失到中间状态的反向传播（输出层）}
\label{fig:9-54}
\end{figure}
......@@ -1756,7 +1756,7 @@ z_t&=&\gamma z_{t-1}+(1-\gamma) \frac{\partial J}{\partial {\theta}_t} \cdot \f
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-back-propagation-output2}
\input{./Chapter9/Figures/figure-back-propagation-output2}
\caption{从中间状态到输入的反向传播（输出层）}
\label{fig:9-55}
\end{figure}
......@@ -1802,7 +1802,7 @@ z_t&=&\gamma z_{t-1}+(1-\gamma) \frac{\partial J}{\partial {\theta}_t} \cdot \f
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-forward-propagation-hid}
\input{./Chapter9/Figures/figure-forward-propagation-hid}
\caption{隐藏层的前向计算过程}
\label{fig:9-56}
\end{figure}
......@@ -1837,7 +1837,7 @@ z_t&=&\gamma z_{t-1}+(1-\gamma) \frac{\partial J}{\partial {\theta}_t} \cdot \f
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-back-propagation-hid}
\input{./Chapter9/Figures/figure-back-propagation-hid}
\caption{隐藏层的反向传播}
\label{fig:9-57}
\end{figure}
......@@ -1902,7 +1902,7 @@ z_t&=&\gamma z_{t-1}+(1-\gamma) \frac{\partial J}{\partial {\theta}_t} \cdot \f
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-4-gram}
\input{./Chapter9/Figures/figure-4-gram}
\caption{4-gram前馈神经网络语言架构}
\label{fig:9-60}
\end{figure}
......@@ -1964,7 +1964,7 @@ z_t&=&\gamma z_{t-1}+(1-\gamma) \frac{\partial J}{\partial {\theta}_t} \cdot \f
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-softmax}
\input{./Chapter9/Figures/figure-softmax}
\caption{ Softmax函数（一维）所对应的曲线}
\label{fig:softmax}
\end{figure}
......@@ -2019,7 +2019,7 @@ z_t&=&\gamma z_{t-1}+(1-\gamma) \frac{\partial J}{\partial {\theta}_t} \cdot \f
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-rnn-lm}
\input{./Chapter9/Figures/figure-rnn-lm}
\caption{基于循环神经网络的语言模型结构}
\label{fig:9-62}
\end{figure}
......@@ -2058,7 +2058,7 @@ z_t&=&\gamma z_{t-1}+(1-\gamma) \frac{\partial J}{\partial {\theta}_t} \cdot \f
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-one-hot}
\input{./Chapter9/Figures/figure-one-hot}
\caption{单词的One-hot表示 }
\label{fig:9-64}
\end{figure}
......@@ -2079,7 +2079,7 @@ z_t&=&\gamma z_{t-1}+(1-\gamma) \frac{\partial J}{\partial {\theta}_t} \cdot \f
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-embedding}
\input{./Chapter9/Figures/figure-embedding}
\caption{单词的分布式表示(词嵌入) }
\label{fig:9-65}
\end{figure}
......@@ -2101,7 +2101,7 @@ z_t&=&\gamma z_{t-1}+(1-\gamma) \frac{\partial J}{\partial {\theta}_t} \cdot \f
%----------------------------------------------
\begin{figure}[htp]
\centering
\includegraphics[width=6cm]{./Chapter9/Figures/word-graph.jpg}
\includegraphics[width=6cm]{./Chapter9/Figures/figure-word-graph.jpg}
\caption{分布式表示的可视化}
\label{fig:9-66}
\end{figure}
......@@ -2112,7 +2112,7 @@ z_t&=&\gamma z_{t-1}+(1-\gamma) \frac{\partial J}{\partial {\theta}_t} \cdot \f
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-embedding-matrix}
\input{./Chapter9/Figures/figure-embedding-matrix}
\caption{词嵌入矩阵${\vectorn{\emph{C}}}$}
\label{fig:9-67}
\end{figure}
......@@ -2143,7 +2143,7 @@ Jobs was the CEO of {\red{\underline{apple}}}.
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-rnn-model}
\input{./Chapter9/Figures/figure-rnn-model}
\caption{基于RNN的表示模型（词+上下文）}
\label{fig:9-68}
\end{figure}
......@@ -2156,7 +2156,7 @@ Jobs was the CEO of {\red{\underline{apple}}}.
%----------------------------------------------
\begin{figure}[htp]
\centering
\input{./Chapter9/Figures/fig-model-training}
\input{./Chapter9/Figures/figure-model-training}
\caption{表示模型的训练方法（与目标任务联合训练 vs 用外部任务预训练）}
\label{fig:9-69}
\end{figure}
......
Markdown 格式
0% 或
您添加了 0 人 到此讨论。请谨慎行事。
请先完成此评论的编辑！
请 注册 或者 后发表评论