Skip to content
项目
群组
代码片段
帮助
当前项目
正在载入...
登录 / 注册
切换导航面板
T
Toy-MT-Introduction
概览
Overview
Details
Activity
Cycle Analytics
版本库
Repository
Files
Commits
Branches
Tags
Contributors
Graph
Compare
Charts
问题
0
Issues
0
列表
Board
标记
里程碑
合并请求
0
Merge Requests
0
CI / CD
CI / CD
流水线
作业
日程表
图表
维基
Wiki
代码片段
Snippets
成员
Collapse sidebar
Close sidebar
活动
图像
聊天
创建新问题
作业
提交
Issue Boards
Open sidebar
NiuTrans
Toy-MT-Introduction
Commits
5ab166f8
Commit
5ab166f8
authored
May 12, 2020
by
xiaotong
Browse files
Options
Browse Files
Download
Plain Diff
Merge branch 'master' of 47.105.50.196:NiuTrans/Toy-MT-Introduction
parents
8468cf3c
beb99716
全部展开
隐藏空白字符变更
内嵌
并排
正在显示
50 个修改的文件
包含
39 行增加
和
57 行删除
+39
-57
Book/Chapter1/Figures/figure-example-NMT.tex
+0
-1
Book/Chapter1/Figures/figure-example-RBMT.tex
+0
-1
Book/Chapter1/Figures/figure-example-SMT.tex
+0
-1
Book/Chapter1/Figures/figure-required-parts-of-MT.tex
+0
-0
Book/Chapter1/chapter1.tex
+1
-1
Book/Chapter2/Figures/figure-example-of-word-segmentation-based-on-dictionary.tex
+0
-1
Book/Chapter2/Figures/figure-probability-density-function&Distribution-function.tex
+0
-1
Book/Chapter2/Figures/figure-self-information-function.tex
+0
-0
Book/Chapter4/chapter4.tex
+38
-12
Book/Chapter6/Chapter6.tex
+0
-0
Book/Chapter6/Figures/figure-a-combination-of-position-encoding-and-word-encoding.tex
+0
-1
Book/Chapter6/Figures/figure-a-working-example-of-neural-machine-translation.tex
+0
-1
Book/Chapter6/Figures/figure-attention-of-source-and-target-words.tex
+0
-1
Book/Chapter6/Figures/figure-automatic-generation-of-ancient-poems-based-on-encoder-decoder-framework.tex
+0
-1
Book/Chapter6/Figures/figure-automatically-generate-instances-of-couplets.tex
+0
-1
Book/Chapter6/Figures/figure-beam-search-process.tex
+0
-1
Book/Chapter6/Figures/figure-calculation-of-context-vector-C.tex
+0
-1
Book/Chapter6/Figures/figure-calculation-process-of-context-vector-C.tex
+0
-1
Book/Chapter6/Figures/figure-comparison-of-the-number-of-padding-in-batch.tex
+0
-1
Book/Chapter6/Figures/figure-data-parallel-process.tex
+0
-1
Book/Chapter6/Figures/figure-decode-the-word-probability-distribution-at-the-first-position.tex
+0
-1
Book/Chapter6/Figures/figure-decoding-process-based-on-greedy-method.tex
+0
-1
Book/Chapter6/Figures/figure-dependencies-between-words-in-a-recurrent-neural-network.tex
+0
-1
Book/Chapter6/Figures/figure-dependencies-between-words-of-attention.tex
+0
-1
Book/Chapter6/Figures/figure-different-regularization-methods.tex
+0
-1
Book/Chapter6/Figures/figure-double-layer-RNN.tex
+0
-1
Book/Chapter6/Figures/figure-example-of-automatic-translation-of-classical-chinese.tex
+0
-1
Book/Chapter6/Figures/figure-example-of-context-vector-calculation-process.tex
+0
-1
Book/Chapter6/Figures/figure-example-of-self-attention-mechanism-calculation.tex
+0
-1
Book/Chapter6/Figures/figure-generate-summary.tex
+0
-1
Book/Chapter6/Figures/figure-mask-instance-for-future-positions-in-transformer.tex
+0
-1
Book/Chapter6/Figures/figure-matrix-representation-of-attention-weights-between-chinese-english-sentence-pairs.tex
+0
-1
Book/Chapter6/Figures/figure-model-structure-based-on-recurrent-neural-network-translation.tex
+0
-1
Book/Chapter6/Figures/figure-multi-head-attention-model.tex
+0
-1
Book/Chapter6/Figures/figure-output-layer-structur.tex
+0
-1
Book/Chapter6/Figures/figure-point-product-attention-model.tex
+0
-1
Book/Chapter6/Figures/figure-position-of-difference-and-layer-regularization-in-the-model.tex
+0
-1
Book/Chapter6/Figures/figure-position-of-feedforward-neural-network-in-the-model.tex
+0
-1
Book/Chapter6/Figures/figure-position-of-self-attention-mechanism-in-the-model.tex
+0
-1
Book/Chapter6/Figures/figure-presentation-space.tex
+0
-0
Book/Chapter6/Figures/figure-query-model-corresponding-to-attention-mechanism.tex
+0
-1
Book/Chapter6/Figures/figure-query-model-corresponding-to-traditional-query-model-vs-attention-mechanism.tex
+0
-1
Book/Chapter6/Figures/figure-query-model-corresponding-to-traditional-query-model-vs-attention-mechanism02.tex
+0
-1
Book/Chapter6/Figures/figure-relationship-between-learning-rate-and-number-of-updates.tex
+0
-1
Book/Chapter6/Figures/figure-residual-network-structure.tex
+0
-1
Book/Chapter6/Figures/figure-structure-of-a-recurrent-network-model.tex
+0
-1
Book/Chapter6/Figures/figure-structure-of-the-network-during-transformer-training.tex
+0
-1
Book/Chapter6/Figures/figure-transformer-input-and-position-encoding.tex
+0
-1
Book/Chapter6/Figures/figure-word-embedding-structure.tex
+0
-1
Book/Chapter7/Figures/figure-underfitting-vs-overfitting.tex
+0
-1
没有找到文件。
Book/Chapter1/Figures/figure-
E
xample-NMT.tex
→
Book/Chapter1/Figures/figure-
e
xample-NMT.tex
查看文件 @
5ab166f8
\definecolor
{
ublue
}{
rgb
}{
0.152,0.250,0.545
}
\definecolor
{
ugreen
}{
rgb
}{
0,0.5,0
}
...
...
Book/Chapter1/Figures/figure-
E
xample-RBMT.tex
→
Book/Chapter1/Figures/figure-
e
xample-RBMT.tex
查看文件 @
5ab166f8
\definecolor
{
ublue
}{
rgb
}{
0.152,0.250,0.545
}
\definecolor
{
ugreen
}{
rgb
}{
0,0.5,0
}
...
...
Book/Chapter1/Figures/figure-
E
xample-SMT.tex
→
Book/Chapter1/Figures/figure-
e
xample-SMT.tex
查看文件 @
5ab166f8
\definecolor
{
ublue
}{
rgb
}{
0.152,0.250,0.545
}
\definecolor
{
ugreen
}{
rgb
}{
0,0.5,0
}
...
...
Book/Chapter1/Figures/figure-
R
equired-parts-of-MT.tex
→
Book/Chapter1/Figures/figure-
r
equired-parts-of-MT.tex
查看文件 @
5ab166f8
File moved
Book/Chapter1/chapter1.tex
查看文件 @
5ab166f8
...
...
@@ -222,7 +222,7 @@
\centering
\input
{
./Chapter1/Figures/figure-comparison-mt-ht-1
}
\end{figure}
\begin{figure}
[
htp
]
\begin{figure}
[
t
]
\centering
\input
{
./Chapter1/Figures/figure-comparison-mt-ht-2
}
\caption
{
机器翻译与人工翻译实例结果对比
}
...
...
Book/Chapter2/Figures/figure-
E
xample-of-word-segmentation-based-on-dictionary.tex
→
Book/Chapter2/Figures/figure-
e
xample-of-word-segmentation-based-on-dictionary.tex
查看文件 @
5ab166f8
\definecolor
{
ublue
}{
rgb
}{
0.152,0.250,0.545
}
\definecolor
{
ugreen
}{
rgb
}{
0,0.5,0
}
...
...
Book/Chapter2/Figures/figure-
P
robability-density-function&Distribution-function.tex
→
Book/Chapter2/Figures/figure-
p
robability-density-function&Distribution-function.tex
查看文件 @
5ab166f8
%%% outline
%-------------------------------------------------------------------------
\begin{tikzpicture}
...
...
Book/Chapter2/Figures/figure-
S
elf-information-function.tex
→
Book/Chapter2/Figures/figure-
s
elf-information-function.tex
查看文件 @
5ab166f8
File moved
Book/Chapter4/chapter4.tex
查看文件 @
5ab166f8
...
...
@@ -1204,7 +1204,7 @@ h_i (d,\textbf{t},\textbf{s})=\sum_{r \in d}h_i (r)
\label
{
eq:4-27
}
\end{eqnarray}
\
parinterval
其中:
\
noindent
其中:
\begin{itemize}
\vspace
{
0.5em
}
...
...
@@ -1430,12 +1430,12 @@ span\textrm{[0,4]}&=&\textrm{``猫} \quad \textrm{喜欢} \quad \textrm{吃} \qu
\parinterval
可以说基于句法的翻译模型贯穿了现代统计机器翻译的发展历程。从概念上讲,不管是层次短语模型,还是语言学句法模型都是基于句法的模型。基于句法的机器翻译模型种类繁多,这里先对相关概念进行简要介绍,以避免后续论述中产生歧义。表
\ref
{
tab:4-2
}
给出了基于句法的机器翻译中涉及的一些概念。
%----------------------------------------------
\begin{table}
[h
t
p]
{
\begin{table}
[h
b
p]
{
\begin{center}
\caption
{
基于句法的机器翻译中常用概念
}
\label
{
tab:4-2
}
{
\begin{tabular}
{
l
| l
}
\begin{tabular}
{
p
{
6.5em
}
| l
}
术语
&
说明
\\
\hline
\rule
{
0pt
}{
15pt
}
翻译规则
&
翻译的最小单元(或步骤)
\\
...
...
@@ -1454,6 +1454,18 @@ span\textrm{[0,4]}&=&\textrm{``猫} \quad \textrm{喜欢} \quad \textrm{吃} \qu
\rule
{
0pt
}{
15pt
}
基于树
&
(源语言)使用树结构(大多指句法树)
\\
\rule
{
0pt
}{
15pt
}
基于串
&
(源语言)使用词串,比如串到树翻译系统的解码器一般
\\
&
都是基于串的解码方法
\\
\end{tabular}
}
\end{center}
}
\end{table}
\vspace
{
3em
}
\begin{table}
[htp]
{
\begin{center}
\vspace
{
1em
}
{
\begin{tabular}
{
p
{
6.5em
}
| l
}
术语
&
说明
\\
\hline
\rule
{
0pt
}{
15pt
}
基于森林
&
(源语言)使用句法森林,这里森林只是对多个句法树的一
\\
&
种压缩表示
\\
\rule
{
0pt
}{
15pt
}
词汇化规则
&
含有终结符的规则
\\
...
...
@@ -1626,7 +1638,7 @@ r_9: \quad \textrm{IP(}\textrm{NN}_1\ \textrm{VP}_2) \rightarrow \textrm{S(}\tex
\end{eqnarray}
}
\
parinterval
可以得到一个翻译推导:
\
noindent
可以得到一个翻译推导:
{
\footnotesize
\begin{eqnarray}
&&
\langle\ \textrm
{
IP
}^{
[1]
}
,
\ \textrm
{
S
}^{
[1]
}
\ \rangle
\nonumber
\\
...
...
@@ -1638,14 +1650,16 @@ r_9: \quad \textrm{IP(}\textrm{NN}_1\ \textrm{VP}_2) \rightarrow \textrm{S(}\tex
&
&
\ \ \textrm
{
S(NP(DT(the) NNS(imports))
}
\ \textrm
{
VP(VBP
}^{
[6]
}
\ \textrm
{
ADVP(RB(drastically)
}
\ \textrm
{
VBN
}^{
[5]
}
)))
\ \rangle
\nonumber
\\
&
\xrightarrow
[r_4]
{
\textrm
{
VV
}^{
[5]
}
\Leftrightarrow
\textrm
{
VBN
}^{
[5]
}}
&
\langle\ \textrm
{
IP(NN(进口)
}
\ \textrm
{
VP(AD(大幅度)
}
\ \textrm
{
VP(VV(减少)
}
\ \textrm
{
AS
}^{
[6]
}
))),
\hspace
{
10em
}
\nonumber
\\
&
&
\ \ \textrm
{
S(NP(DT(the) NNS(imports))
}
\ \textrm
{
VP(VBP
}^{
[6]
}
\ \nonumber
\\
&
&
\ \ \textrm
{
ADVP(RB(drastically)
}
\ \textrm
{
VBN(fallen)
}
)))
\ \rangle
\nonumber
\\
&
&
\ \ \textrm
{
ADVP(RB(drastically)
}
\ \textrm
{
VBN(fallen)
}
)))
\ \rangle
\nonumber
\end{eqnarray}
\begin{eqnarray}
&
\xrightarrow
[r_6]
{
\textrm
{
AS
}^{
[6]
}
\Leftrightarrow
\textrm
{
VBP
}^{
[6]
}}
&
\langle\ \textrm
{
IP(NN(进口)
}
\ \textrm
{
VP(AD(大幅度)
}
\ \textrm
{
VP(VV(减少)
}
\ \textrm
{
AS(了)
}
))),
\nonumber
\\
&
&
\ \ \textrm
{
S(NP(DT(the) NNS(imports))
}
\ \textrm
{
VP(VBP(have)
}
\ \nonumber
\\
&
&
\ \ \textrm
{
ADVP(RB(drastically)
}
\ \textrm
{
VBN(fallen)
}
)))
\ \rangle
\hspace
{
15em
}
\nonumber
\end{eqnarray}
}
\
parinterval
其中,箭头
$
\rightarrow
$
表示推导之意。显然,可以把翻译看作是基于树结构的推导过程(记为
$
d
$
)。因此,与层次短语模型一样,基于语言学句法的机器翻译也是要找到最佳的推导
$
\hat
{
d
}
=
\arg\max\textrm
{
P
}
(
d
)
$
。
\
noindent
其中,箭头
$
\rightarrow
$
表示推导之意。显然,可以把翻译看作是基于树结构的推导过程(记为
$
d
$
)。因此,与层次短语模型一样,基于语言学句法的机器翻译也是要找到最佳的推导
$
\hat
{
d
}
=
\arg\max\textrm
{
P
}
(
d
)
$
。
%----------------------------------------------------------------------------------------
% NEW SUBSUB-SECTION
...
...
@@ -1664,7 +1678,7 @@ r_9: \quad \textrm{IP(}\textrm{NN}_1\ \textrm{VP}_2) \rightarrow \textrm{S(}\tex
\end{figure}
%-------------------------------------------
\
parinterval
其中,源语言树片段中的叶子结点NN表示变量,它与右手端的变量NN对应。这里仍然可以使用基于树结构的规则对上面这个树到串的映射进行表示。参照规则形式
$
\langle\ \alpha
_
h,
\beta
_
h
\ \rangle
\to
\langle\ \alpha
_
r,
\beta
_
r,
\sim\ \rangle
$
,有:
\
noindent
其中,源语言树片段中的叶子结点NN表示变量,它与右手端的变量NN对应。这里仍然可以使用基于树结构的规则对上面这个树到串的映射进行表示。参照规则形式
$
\langle\ \alpha
_
h,
\beta
_
h
\ \rangle
\to
\langle\ \alpha
_
r,
\beta
_
r,
\sim\ \rangle
$
,有:
\begin{eqnarray}
\alpha
_
h
&
=
&
\textrm
{
VP
}
\nonumber
\\
\beta
_
h
&
=
&
\textrm
{
VP
}
\
(=
\alpha
_
h)
\nonumber
\\
...
...
@@ -1800,7 +1814,7 @@ r_9: \quad \textrm{IP(}\textrm{NN}_1\ \textrm{VP}_2) \rightarrow \textrm{S(}\tex
\textrm
{
VP(PP(P(对)
}
\ \textrm
{
NP(NN(回答)))
}
\ \textrm
{
VP
}_
1)
\rightarrow
\textrm
{
VP
}_
1
\ \textrm
{
with
}
\ \textrm
{
the
}
\ \textrm
{
answer
}
\nonumber
\end{eqnarray}
\
parinterval
其中,蓝色部分表示可以抽取到的规则,显然它的根节点和叶子非终结符节点都是可信节点。由于源语言树片段中包含一个变量(VP),因此需要对VP节点的Span所表示的目标语言范围进行泛化(红色方框部分)。
\
noindent
其中,蓝色部分表示可以抽取到的规则,显然它的根节点和叶子非终结符节点都是可信节点。由于源语言树片段中包含一个变量(VP),因此需要对VP节点的Span所表示的目标语言范围进行泛化(红色方框部分)。
%----------------------------------------------
\begin{figure}
[htp]
...
...
@@ -1985,7 +1999,7 @@ r_9: \quad \textrm{IP(}\textrm{NN}_1\ \textrm{VP}_2) \rightarrow \textrm{S(}\tex
\textrm
{
VP(
}
\textrm
{
PP
}_
1
\ \textrm
{
VP(VV(表示)
}
\ \textrm
{
NN
}_
2
\textrm
{
))
}
\rightarrow
\textrm
{
VP(VBZ(was)
}
\ \textrm
{
VP(
}
\textrm
{
VBN
}_
2
\ \textrm
{
PP
}_
1
\textrm
{
))
}
\nonumber
\end{eqnarray}
\
parinterval
其中,规则的左部是源语言句法树结构,右部是目标语言句法树结构,变量的下标表示对应关系。为了获取这样的规则,需要进行树到树规则抽取。最直接的办法是把GHKM方法推广到树到树翻译的情况。比如,可以利用双语结构的约束和词对齐,定义树的切割点,之后找到两种语言树结构的映射关系
\cite
{
liu2009improving
}
。
\
noindent
其中,规则的左部是源语言句法树结构,右部是目标语言句法树结构,变量的下标表示对应关系。为了获取这样的规则,需要进行树到树规则抽取。最直接的办法是把GHKM方法推广到树到树翻译的情况。比如,可以利用双语结构的约束和词对齐,定义树的切割点,之后找到两种语言树结构的映射关系
\cite
{
liu2009improving
}
。
%----------------------------------------------------------------------------------------
% NEW SUBSUB-SECTION
...
...
@@ -2007,7 +2021,7 @@ r_9: \quad \textrm{IP(}\textrm{NN}_1\ \textrm{VP}_2) \rightarrow \textrm{S(}\tex
\parinterval
换一个角度来看,词对齐实际上只是帮助模型找到两种语言句法树中节点的对应关系。如果能够直接得到句法树节点的对应,就可以避免掉词对齐的错误。也就是,可以直接使用节点对齐来进行树到树规则的抽取。首先,利用外部的节点对齐工具获得两棵句法树节点之间的对齐关系。之后,将每个对齐的节点看作是树片段的根节点,再进行规则抽取。图
\ref
{
fig:4-62
}
展示了基于节点对齐的规则抽取结果。
%----------------------------------------------
\begin{figure}
[ht
p
]
\begin{figure}
[ht
b
]
\centering
\input
{
./Chapter4/Figures/tree-to-tree-rule-extraction-base-node-alignment
}
\caption
{
基于节点对齐的树到树规则抽取
}
...
...
@@ -2205,12 +2219,24 @@ d_1 = {d'} \circ {r_5}
\caption
{
基于串的解码 vs 基于树的解码
}
\label
{
tab:4-4
}
{
\begin{tabular}
{
l |
l
l
}
\begin{tabular}
{
l |
p
{
16.5em
}
l
}
对比
&
基于树的解码
&
基于串的解码
\\
\hline
\rule
{
0pt
}{
15pt
}
解码方法
&
$
\hat
{
d
}
=
\arg\max
_{
d
\in
D
_{
\textrm
{
tree
}}}
\textrm
{
score
}
(
d
)
$
&
$
\hat
{
d
}
=
\arg\max
_{
d
\in
D
}
\textrm
{
score
}
(
d
)
$
\\
\rule
{
0pt
}{
15pt
}
搜索空间
&
与输入的源语句法树兼容的推导
$
D
_{
\textrm
{
tree
}}$
&
所有的推导
$
D
$
\\
\rule
{
0pt
}{
15pt
}
适用模型
&
树到串、树到树
&
所有的句法模型
\\
\rule
{
0pt
}{
15pt
}
适用模型
&
树到串、树到树
&
所有的句法模型
\end{tabular}
}
\end{center}
}
\end{table}
\begin{table}
[htp]
{
\begin{center}
\vspace
{
1em
}
{
\begin{tabular}
{
l | p
{
16.5em
}
l
}
对比
&
基于树的解码
&
基于串的解码
\\
\hline
\rule
{
0pt
}{
15pt
}
解码算法
&
Chart解码
&
CKY + 规则二叉化
\\
\rule
{
0pt
}{
15pt
}
速度
&
快
&
一般较慢
\end{tabular}
...
...
Book/Chapter6/Chapter6.tex
查看文件 @
5ab166f8
差异被折叠。
点击展开。
Book/Chapter6/Figures/figure-
A
-combination-of-position-encoding-and-word-encoding.tex
→
Book/Chapter6/Figures/figure-
a
-combination-of-position-encoding-and-word-encoding.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\begin{scope}
\tikzstyle
{
rnode
}
= [draw,minimum width=3.5em,minimum height=1.2em]
...
...
Book/Chapter6/Figures/figure-
A
-working-example-of-neural-machine-translation.tex
→
Book/Chapter6/Figures/figure-
a
-working-example-of-neural-machine-translation.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\setlength
{
\base
}{
1cm
}
...
...
Book/Chapter6/Figures/figure-
A
ttention-of-source-and-target-words.tex
→
Book/Chapter6/Figures/figure-
a
ttention-of-source-and-target-words.tex
查看文件 @
5ab166f8
%
%---------------------------------------
...
...
Book/Chapter6/Figures/figure-
A
utomatic-generation-of-ancient-poems-based-on-encoder-decoder-framework.tex
→
Book/Chapter6/Figures/figure-
a
utomatic-generation-of-ancient-poems-based-on-encoder-decoder-framework.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\begin{scope}
\tikzstyle
{
lnode
}
= [minimum height=2em,minimum width=8em,inner sep=3pt,rounded corners=2pt,draw,fill=red!20];
...
...
Book/Chapter6/Figures/figure-
A
utomatically-generate-instances-of-couplets.tex
→
Book/Chapter6/Figures/figure-
a
utomatically-generate-instances-of-couplets.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\begin{scope}
\tikzstyle
{
lnode
}
= [minimum height=2.5em,minimum width=12em,inner sep=3pt,rounded corners=2pt,draw=red!75!black,fill=red!5];
...
...
Book/Chapter6/Figures/figure-
B
eam-search-process.tex
→
Book/Chapter6/Figures/figure-
b
eam-search-process.tex
查看文件 @
5ab166f8
%-----------------------------------------
...
...
Book/Chapter6/Figures/figure-
C
alculation-of-context-vector-C.tex
→
Book/Chapter6/Figures/figure-
c
alculation-of-context-vector-C.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\begin{scope}
...
...
Book/Chapter6/Figures/figure-
C
alculation-process-of-context-vector-C.tex
→
Book/Chapter6/Figures/figure-
c
alculation-process-of-context-vector-C.tex
查看文件 @
5ab166f8
...
...
@@ -4,7 +4,6 @@
\begin{tikzpicture}
\begin{scope}
...
...
Book/Chapter6/Figures/figure-
C
omparison-of-the-number-of-padding-in-batch.tex
→
Book/Chapter6/Figures/figure-
c
omparison-of-the-number-of-padding-in-batch.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\begin{scope}
[scale=1.5]
...
...
Book/Chapter6/Figures/figure-
D
ata-parallel-process.tex
→
Book/Chapter6/Figures/figure-
d
ata-parallel-process.tex
查看文件 @
5ab166f8
%----------------------------------
...
...
Book/Chapter6/Figures/figure-
D
ecode-the-word-probability-distribution-at-the-first-position.tex
→
Book/Chapter6/Figures/figure-
d
ecode-the-word-probability-distribution-at-the-first-position.tex
查看文件 @
5ab166f8
...
...
@@ -3,7 +3,6 @@
\begin{tikzpicture}
\begin{scope}
\tikzstyle
{
rnnnode
}
= [minimum height=1.1em,minimum width=3.5em,inner sep=2pt,rounded corners=1pt,draw,fill=red!20];
...
...
Book/Chapter6/Figures/figure-
D
ecoding-process-based-on-greedy-method.tex
→
Book/Chapter6/Figures/figure-
d
ecoding-process-based-on-greedy-method.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\begin{scope}
\tikzstyle
{
rnnnode
}
= [minimum height=1.1em,minimum width=2.1em,inner sep=2pt,rounded corners=1pt,draw,fill=red!20];
...
...
Book/Chapter6/Figures/figure-
D
ependencies-between-words-in-a-recurrent-neural-network.tex
→
Book/Chapter6/Figures/figure-
d
ependencies-between-words-in-a-recurrent-neural-network.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\begin{scope}
\node
[anchor=west] (w0) at (0,0)
{$
w
_
1
$}
;
...
...
Book/Chapter6/Figures/figure-
Dependencies-between-words-of-A
ttention.tex
→
Book/Chapter6/Figures/figure-
dependencies-between-words-of-a
ttention.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\begin{scope}
\node
[anchor=west] (w0) at (0,-2)
{$
w
_
1
$}
;
...
...
Book/Chapter6/Figures/figure-
D
ifferent-regularization-methods.tex
→
Book/Chapter6/Figures/figure-
d
ifferent-regularization-methods.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\begin{scope}
\tikzstyle
{
lnode
}
= [minimum height=1.5em,minimum width=3em,inner sep=3pt,rounded corners=1.5pt,draw,fill=orange!20];
...
...
Book/Chapter6/Figures/figure-
D
ouble-layer-RNN.tex
→
Book/Chapter6/Figures/figure-
d
ouble-layer-RNN.tex
查看文件 @
5ab166f8
%--------------------------------------------------------------------------------
\begin{tikzpicture}
\setlength
{
\base
}{
0.9cm
}
...
...
Book/Chapter6/Figures/figure-
Example-of-automatic-translation-of-classical-C
hinese.tex
→
Book/Chapter6/Figures/figure-
example-of-automatic-translation-of-classical-c
hinese.tex
查看文件 @
5ab166f8
%---------------------------------------
\begin{frame}
{}
...
...
Book/Chapter6/Figures/figure-
E
xample-of-context-vector-calculation-process.tex
→
Book/Chapter6/Figures/figure-
e
xample-of-context-vector-calculation-process.tex
查看文件 @
5ab166f8
%-------------------------------------------
...
...
Book/Chapter6/Figures/figure-
E
xample-of-self-attention-mechanism-calculation.tex
→
Book/Chapter6/Figures/figure-
e
xample-of-self-attention-mechanism-calculation.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\begin{scope}
...
...
Book/Chapter6/Figures/figure-
G
enerate-summary.tex
→
Book/Chapter6/Figures/figure-
g
enerate-summary.tex
查看文件 @
5ab166f8
%%% outline
%-------------------------------------------------------------------------
\begin{tikzpicture}
...
...
Book/Chapter6/Figures/figure-
Mask-instance-for-future-positions-in-T
ransformer.tex
→
Book/Chapter6/Figures/figure-
mask-instance-for-future-positions-in-t
ransformer.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\begin{scope}
\tikzstyle
{
attnode
}
= [minimum size=1.5em,inner sep=0pt,rounded corners=1pt,draw]
...
...
Book/Chapter6/Figures/figure-
Matrix-Representation-of-Attention-Weights-Between-Chinese-English-Sentence-P
airs.tex
→
Book/Chapter6/Figures/figure-
matrix-representation-of-attention-weights-between-chinese-english-sentence-p
airs.tex
查看文件 @
5ab166f8
...
...
@@ -3,7 +3,6 @@
%-------------------------------------------
\begin{tikzpicture}
...
...
Book/Chapter6/Figures/figure-
M
odel-structure-based-on-recurrent-neural-network-translation.tex
→
Book/Chapter6/Figures/figure-
m
odel-structure-based-on-recurrent-neural-network-translation.tex
查看文件 @
5ab166f8
%--------------------------------------------------------------------------
\begin{tikzpicture}
...
...
Book/Chapter6/Figures/figure-
Multi-Head-Attention-M
odel.tex
→
Book/Chapter6/Figures/figure-
multi-head-attention-m
odel.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\begin{scope}
...
...
Book/Chapter6/Figures/figure-
O
utput-layer-structur.tex
→
Book/Chapter6/Figures/figure-
o
utput-layer-structur.tex
查看文件 @
5ab166f8
% not compatible with [scale=?]
...
...
Book/Chapter6/Figures/figure-
P
oint-product-attention-model.tex
→
Book/Chapter6/Figures/figure-
p
oint-product-attention-model.tex
查看文件 @
5ab166f8
...
...
@@ -3,7 +3,6 @@
\begin{tikzpicture}
\begin{scope}
...
...
Book/Chapter6/Figures/figure-
P
osition-of-difference-and-layer-regularization-in-the-model.tex
→
Book/Chapter6/Figures/figure-
p
osition-of-difference-and-layer-regularization-in-the-model.tex
查看文件 @
5ab166f8
...
...
@@ -2,7 +2,6 @@
\begin{tikzpicture}
\begin{scope}
\tikzstyle
{
Sanode
}
= [minimum height=1.4em,minimum width=7em,inner sep=3pt,rounded corners=1.5pt,draw,fill=orange!20];
...
...
Book/Chapter6/Figures/figure-
P
osition-of-feedforward-neural-network-in-the-model.tex
→
Book/Chapter6/Figures/figure-
p
osition-of-feedforward-neural-network-in-the-model.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\begin{scope}
\tikzstyle
{
Sanode
}
= [minimum height=1.4em,minimum width=7em,inner sep=3pt,rounded corners=1.5pt,draw,fill=orange!20];
...
...
Book/Chapter6/Figures/figure-
P
osition-of-self-attention-mechanism-in-the-model.tex
→
Book/Chapter6/Figures/figure-
p
osition-of-self-attention-mechanism-in-the-model.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\begin{scope}
\tikzstyle
{
Sanode
}
= [minimum height=1.4em,minimum width=7em,inner sep=3pt,rounded corners=1.5pt,draw,fill=orange!20];
...
...
Book/Chapter6/Figures/figure-
P
resentation-space.tex
→
Book/Chapter6/Figures/figure-
p
resentation-space.tex
查看文件 @
5ab166f8
File moved
Book/Chapter6/Figures/figure-
Q
uery-model-corresponding-to-attention-mechanism.tex
→
Book/Chapter6/Figures/figure-
q
uery-model-corresponding-to-attention-mechanism.tex
查看文件 @
5ab166f8
...
...
@@ -7,7 +7,6 @@
\begin{tikzpicture}
\begin{scope}
...
...
Book/Chapter6/Figures/figure-
Q
uery-model-corresponding-to-traditional-query-model-vs-attention-mechanism.tex
→
Book/Chapter6/Figures/figure-
q
uery-model-corresponding-to-traditional-query-model-vs-attention-mechanism.tex
查看文件 @
5ab166f8
%-----------------------------------------------------
\begin{tikzpicture}
...
...
Book/Chapter6/Figures/figure-
Q
uery-model-corresponding-to-traditional-query-model-vs-attention-mechanism02.tex
→
Book/Chapter6/Figures/figure-
q
uery-model-corresponding-to-traditional-query-model-vs-attention-mechanism02.tex
查看文件 @
5ab166f8
%-----------------------------------------------------
\begin{tikzpicture}
...
...
Book/Chapter6/Figures/figure-
R
elationship-between-learning-rate-and-number-of-updates.tex
→
Book/Chapter6/Figures/figure-
r
elationship-between-learning-rate-and-number-of-updates.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\footnotesize
{
\begin{axis}
[
...
...
Book/Chapter6/Figures/figure-
R
esidual-network-structure.tex
→
Book/Chapter6/Figures/figure-
r
esidual-network-structure.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\begin{scope}
\tikzstyle
{
lnode
}
= [minimum height=1.5em,minimum width=3em,inner sep=3pt,rounded corners=1.5pt,draw,fill=orange!20];
...
...
Book/Chapter6/Figures/figure-
S
tructure-of-a-recurrent-network-model.tex
→
Book/Chapter6/Figures/figure-
s
tructure-of-a-recurrent-network-model.tex
查看文件 @
5ab166f8
%------------------------------------------------------------
\begin{tikzpicture}
...
...
Book/Chapter6/Figures/figure-
Structure-of-the-network-during-T
ransformer-training.tex
→
Book/Chapter6/Figures/figure-
structure-of-the-network-during-t
ransformer-training.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\begin{scope}
\tikzstyle
{
rnnnode
}
= [minimum height=1.1em,minimum width=2.1em,inner sep=2pt,rounded corners=1pt,draw,fill=red!20];
...
...
Book/Chapter6/Figures/figure-
T
ransformer-input-and-position-encoding.tex
→
Book/Chapter6/Figures/figure-
t
ransformer-input-and-position-encoding.tex
查看文件 @
5ab166f8
\begin{tikzpicture}
\begin{scope}
\tikzstyle
{
Sanode
}
= [minimum height=1.4em,minimum width=7em,inner sep=3pt,rounded corners=1.5pt,draw];
...
...
Book/Chapter6/Figures/figure-
W
ord-embedding-structure.tex
→
Book/Chapter6/Figures/figure-
w
ord-embedding-structure.tex
查看文件 @
5ab166f8
%------------------------------------------------------
\begin{tikzpicture}
...
...
Book/Chapter7/Figures/figure-
Underfitting-vs-O
verfitting.tex
→
Book/Chapter7/Figures/figure-
underfitting-vs-o
verfitting.tex
查看文件 @
5ab166f8
\begin{tabular}
{
l l l
}
\begin{tikzpicture}
\draw
[->, thick]
(0,0) to (3,0);
...
...
编写
预览
Markdown
格式
0%
重试
或
添加新文件
添加附件
取消
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!
取消
请
注册
或者
登录
后发表评论