Skip to content

T
Toy-MT-Introduction
Commit cccce9da by xiaotong
Options

new pages

parent bea44fa2
正在显示 包含 434 行增加 和 0 行删除
Section06-Neural-Machine-Translation/section06-test.tex 0 → 100644
% !Mode:: "TeX:UTF-8"
% !TEX encoding = UTF-8 Unicode
\def\CTeXPreproc{Created by ctex v0.2.13, don't edit!}
\documentclass[cjk,t,compress,12pt]{beamer}
\usepackage{pstricks}
\usepackage{etex}
\usepackage{eso-pic,graphicx}
\usepackage{fancybox}
\usepackage{amsmath,amssymb}
\usepackage{setspace}
\usepackage{xcolor}
\usepackage{array,multirow}
\usepackage{CJK}
\usepackage{tikz}
\usepackage{tikz-qtree}
\usepackage{hyperref}
\usepackage{changepage}
\usepackage{pgfplots}
\usepackage{subfigure}
\usepackage{tikz-3dplot}
\usepackage{esvect}
\usepackage{CJKulem}
\usepackage{tcolorbox}
\tcbuselibrary{skins}
\usetikzlibrary{calc,intersections}
\usetikzlibrary{matrix}
\usetikzlibrary{arrows,decorations.pathreplacing}
\usetikzlibrary{shadows} % LATEX and plain TEX when using Tik Z
\usetikzlibrary{shadows.blur}
\usepgflibrary{arrows} % LATEX and plain TEX and pure pgf
\usetikzlibrary{arrows} % LATEX and plain TEX when using Tik Z
\usetikzlibrary{decorations}
\usetikzlibrary{arrows,shapes}
\usetikzlibrary{positioning,fit,calc}
\usetikzlibrary{mindmap,backgrounds} % mind map
\DeclareMathOperator*{\argmax}{arg\,max}
\DeclareMathOperator*{\argmin}{arg\,min}
\setbeamertemplate{items}[ball]
\usefonttheme[onlymath]{serif} % fout of math
\definecolor{ugreen}{rgb}{0,0.5,0}
\definecolor{lgreen}{rgb}{0.9,1,0.8}
\definecolor{xtgreen1}{rgb}{0.824,0.898,0.8}
\definecolor{xtgreen}{rgb}{0.914,0.945,0.902}
\definecolor{lightgray}{gray}{0.85}
\setbeamercolor{uppercol}{fg=white,bg=ugreen}
\setbeamercolor{lowercol}{fg=black,bg=xtgreen}
\definecolor{ublue}{rgb}{0.152,0.250,0.545}
\setbeamercolor{uppercolblue}{fg=white,bg=ublue}
\setbeamercolor{lowercolblue}{fg=black,bg=blue!10}
%\usetheme{default}
%\usetheme{Darmstadt}
%\usetheme{Madrid}
%\usetheme{Frankfurt}
%\usetheme{Dresden}
%\usetheme{Boadilla}
%\usecolortheme{dolphin}
\newcounter{mycount1}
\newcounter{mycount2}
\newcounter{mycount3}
\newcounter{mycount4}
\usefonttheme[onlylarge]{structurebold}
\IfFileExists{C:/WINDOWS/win.ini}
{\newcommand{\mycfont}{you}}
{\newcommand{\mycfont}{gbsn}}
\begin{CJK}{UTF8}{\mycfont}
\end{CJK}
\setbeamerfont*{frametitle}{size=\large,series=\bfseries}
\setbeamertemplate{navigation symbols}{\begin{CJK}{UTF8}{\mycfont} 第六章 神经机器翻译 \hspace*{2em} 肖桐\&朱靖波 \end{CJK} \hspace*{2em} \today \hspace*{2em} \insertframenumber{}/\inserttotalframenumber}
\setbeamertemplate{itemize items}[circle] % if you want a circle
\setbeamertemplate{itemize subitem}[triangle] % if you wnat a triangle
\setbeamertemplate{itemize subsubitem}[ball] % if you want a ball
\begin{document}
\begin{CJK}{UTF8}{\mycfont}
\title{\Large{神经机器翻译}}
\author{\large{\textbf{肖桐\ \ 朱靖波}}}
\institute{
\blue{\url{xiaotong@mail.neu.edu.cn}} \black{} \\
\blue{\url{zhujingbo@mail.neu.edu.cn}} \black{} \\
\vspace{1.0em}
东北大学 自然语言处理实验室 \\
\blue{\underline{\url{http://www.nlplab.com}}} \black{} \\
\vspace{0.2cm}
\hspace{0.1cm} \includegraphics[scale=0.1]{../Figures/logo.pdf}
}
\date{}
\maketitle
\setlength{\leftmargini}{1em}
\setlength{\leftmarginii}{1em}
%%%------------------------------------------------------------------------------------------------------------
\section{编码器-解码器框架}
%%%------------------------------------------------------------------------------------------------------------
%%% 译文对比
\begin{frame}{今天机器翻译的水平}
\begin{tcolorbox}[enhanced,frame engine=empty,boxrule=0.1mm,size=title,colback=red!10!white]
\begin{flushleft}
\begin{spacing}{0.9}
\scriptsize{\textbf{原文(英语)}: During Soviet times, if a city's population topped one million, it would become eligible for its own metro. Planners wanted to brighten the lives of everyday Soviet citizens, and saw the metros, with their tens of thousands of daily passengers, as a singular opportunity to do so. In 1977, Tashkent, the capital of Uzbekistan, became the seventh Soviet city to have a metro built. Grand themes celebrating the history of Uzbekistan and the Soviet Union were brought to life, as art was commissioned and designers set to work. The stations reflected different themes, some with domed ceilings and painted tiles reminiscent of Uzbekistan's Silk Road mosques, while others ... }
\end{spacing}
\end{flushleft}
\end{tcolorbox}
\begin{minipage}[t]{0.47\textwidth}
\begin{beamerboxesrounded}[upper=uppercolblue,lower=lowercolblue,shadow=true]{\scriptsize{\textbf{译文\visible<2->{(统计机器翻译)}}}}
{\scriptsize
\begin{spacing}{0.9}
在苏联时代，如果一个城市的人口突破一百万，这将成为合资格为自己的地铁。规划者想去照亮每天的苏联公民的生命，看到地铁，与他们的数十每天数千乘客，作为一个独特的机会来这样做。1977年，塔什干，乌兹别克斯坦的首都，成了苏联第七城市建有地铁。宏大主题，庆祝乌兹别克斯坦和苏联的历史被带到生活，因为艺术是委托和设计师开始工作。车站反映了不同的主题，有的圆顶天花板和绘瓷砖让人想起乌兹别克斯坦是丝绸之路的清真寺，而另一些则装饰着...
\end{spacing}
}
\end{beamerboxesrounded}
\end{minipage}
\hfill
\begin{minipage}[t]{0.47\textwidth}
\begin{beamerboxesrounded}[upper=uppercolblue,lower=lowercolblue,shadow=true]{\scriptsize{\textbf{译文\visible<2->{(神经机器翻译 - 很流畅！)}}}}
{\scriptsize
\begin{spacing}{0.9}
在苏联时期，如果一个城市的人口超过一百万，它就有资格拥有自己的地铁。 规划者想要照亮日常苏联公民的生活，并把拥有数万名每日乘客的地铁看作是这样做的一个绝佳机会。 1977年，乌兹别克斯坦首都塔什干成为苏联第七个修建地铁的城市。 随着艺术的委托和设计师们的工作，乌兹别克斯坦和苏联历史的宏伟主题被赋予了生命力。 这些电台反映了不同的主题，有的有穹顶和彩砖，让人想起乌兹别克斯坦的丝绸之路清真寺，有的则用...
\end{spacing}
}
\end{beamerboxesrounded}
\end{minipage}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
\section{循环神经网络翻译模型及注意力机制}
%%%------------------------------------------------------------------------------------------------------------
\subsection{模型结构}
%%%------------------------------------------------------------------------------------------------------------
\subsection{注意力机制}
%%%------------------------------------------------------------------------------------------------------------
\section{Transformer}
%%%------------------------------------------------------------------------------------------------------------
\subsection{多头自注意力模型}
%%%------------------------------------------------------------------------------------------------------------
\subsection{训练}
%%%------------------------------------------------------------------------------------------------------------
\subsection{推断}
%%%------------------------------------------------------------------------------------------------------------
\subsection{深层网络}
%%%------------------------------------------------------------------------------------------------------------
\section{其它应用}
\end{CJK}
\end{document}
Section06-Neural-Machine-Translation/section06.tex
......@@ -20,6 +20,7 @@
\usepackage{subfigure}
\usepackage{tikz-3dplot}
\usepackage{esvect}
\usepackage{CJKulem}
\usepackage{tcolorbox}
\tcbuselibrary{skins}
......@@ -114,6 +115,255 @@
\section{编码器-解码器框架}
%%%------------------------------------------------------------------------------------------------------------
%%% landscape
\begin{frame}{Landscape}
\vspace{-1em}
\begin{center}
\begin{tikzpicture}
\node [] (history) at (0,0) {\includegraphics[scale=0.29]{./Figures/mt-history.png}};
\draw[-,thick] ([xshift=-1.1in]history.south) coordinate (top01) -- ([yshift=-2em]top01);
\draw[-,thick] ([xshift=1in]history.south) coordinate (top02) -- ([yshift=-2em]top02);
\draw[-,thick] ([xshift=-0.3in]history.south east) coordinate (top03) -- ([yshift=-2em]top03);
\draw[<->,thick] ([yshift=-1em]top01) -- ([yshift=-1em]top02) node [pos=0.5,below,align=left,yshift=-0.5em] {\footnotesize{\textbf{Sections 3-4:}}\\\footnotesize{统计机器翻译}};
\draw[<->,thick] ([yshift=-1em]top02) -- ([yshift=-1em]top03) node [pos=0.5,below,align=left,yshift=-0.5em] {\footnotesize{\textbf{Sections 5-6:}}\\\footnotesize{神经机器翻译}};
\visible<2->{
\draw[<-,thick] ([xshift=0.9in,yshift=-0.2in]history.north) coordinate (label01) -- ([yshift=1.5em,xshift=-1.5em]label01) coordinate (label02);
\node[anchor=south west,align=left] (label) at ([xshift=-5em]label02) {\textbf{\alert{本章内容（Section 6）：}}\\神经机器翻译建模及实现};
}
\end{tikzpicture}
\end{center}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 译文对比
\begin{frame}{今天机器翻译的水平}
\begin{tcolorbox}[enhanced,frame engine=empty,boxrule=0.1mm,size=title,colback=red!10!white]
\begin{flushleft}
\begin{spacing}{0.9}
\scriptsize{\textbf{原文(英语)}: During Soviet times, if a city's population topped one million, it would become eligible for its own metro. Planners wanted to brighten the lives of everyday Soviet citizens, and saw the metros, with their tens of thousands of daily passengers, as a singular opportunity to do so. In 1977, Tashkent, the capital of Uzbekistan, became the seventh Soviet city to have a metro built. Grand themes celebrating the history of Uzbekistan and the Soviet Union were brought to life, as art was commissioned and designers set to work. The stations reflected different themes, some with domed ceilings and painted tiles reminiscent of Uzbekistan's Silk Road mosques, while others ... }
\end{spacing}
\end{flushleft}
\end{tcolorbox}
\begin{minipage}[t]{0.47\textwidth}
\begin{beamerboxesrounded}[upper=uppercolblue,lower=lowercolblue,shadow=true]{\scriptsize{\textbf{译文\visible<2->{(统计机器翻译)}}}}
{\scriptsize
\begin{spacing}{0.9}
在苏联时代，如果一个城市的人口突破一百万，这将成为合资格为自己的地铁。规划者想去照亮每天的苏联公民的生命，看到地铁，与他们的数十每天数千乘客，作为一个独特的机会来这样做。1977年，塔什干，乌兹别克斯坦的首都，成了苏联第七城市建有地铁。宏大主题，庆祝乌兹别克斯坦和苏联的历史被带到生活，因为艺术是委托和设计师开始工作。车站反映了不同的主题，有的圆顶天花板和绘瓷砖让人想起乌兹别克斯坦是丝绸之路的清真寺，而另一些则装饰着...
\end{spacing}
}
\end{beamerboxesrounded}
\end{minipage}
\hfill
\begin{minipage}[t]{0.47\textwidth}
\begin{beamerboxesrounded}[upper=uppercolblue,lower=lowercolblue,shadow=true]{\scriptsize{\textbf{译文\visible<2->{(神经机器翻译 - 很流畅！)}}}}
{\scriptsize
\begin{spacing}{0.9}
在苏联时期，如果一个城市的人口超过一百万，它就有资格拥有自己的地铁。 规划者想要照亮日常苏联公民的生活，并把拥有数万名每日乘客的地铁看作是这样做的一个绝佳机会。 1977年，乌兹别克斯坦首都塔什干成为苏联第七个修建地铁的城市。 随着艺术的委托和设计师们的工作，乌兹别克斯坦和苏联历史的宏伟主题被赋予了生命力。 这些电台反映了不同的主题，有的有穹顶和彩砖，让人想起乌兹别克斯坦的丝绸之路清真寺，有的则用...
\end{spacing}
}
\end{beamerboxesrounded}
\end{minipage}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 神经机器翻译的性能增长
\begin{frame}{神经机器翻译的进展}
\begin{itemize}
\item 想当年，08年NIST举办的汉英机器翻译评测，BLEU能突破30已经是巨猛无比的结果，而现在的神经机器翻译轻松突破45！
\item WMT的结果？？
\end{itemize}
%%% 秀smt nmt的blue随着年代的变化
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 神经机器翻译的性能增长
\begin{frame}{神经机器翻译的进展(续)}
\begin{itemize}
\item 性能提升
\item Human party MT
\end{itemize}
%%% 进一步找一些证据说明NMT的进展，很大google、microsoft的文章，可以看看提升了多少性能等等
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 传统方法 vs 神经网络方法
\begin{frame}{神经机器翻译的优势}
\begin{itemize}
\item 神经网络方法带来了新的思路，还是同样的表再看一遍
\end{itemize}
\begin{tabular} {l | l}
\textbf{传统基于统计的方法} & \textbf{深度学习方法} \\
\textbf{(统计机器翻译)} & \textbf{(神经机器翻译)} \\ \hline
基于离散空间的表示模型 & 基于连续空间的表示模型 \\
NLP问题的隐含结构假设 & 无隐含结构假设，端到端学习 \\
特征工程为主 & 无显性特征，但需要设计网络 \\
特征、规则的存储耗资源 & 模型存储相对小，但计算慢
\end{tabular}
\begin{itemize}
\item<2-> \only<2>{在统计机器翻译时代，系统依赖很多模块，比如}\only<3->{\sout{在统计机器翻译时代，系统依赖很多模块，比如}}
\begin{itemize}
\item \only<2>{\textbf{词对齐}：双语句子之间词和词的对应关系}\only<3->{\sout{\textbf{词对齐}：双语句子之间词和词的对应关系}}
\item \only<2>{\textbf{短语(规则)表}：原文和译文互译的片段}\only<3->{\sout{\textbf{短语(规则)表}：原文和译文互译的片段}}
\item \only<2>{\textbf{特征}：人工设计的用于建模翻译的各种各样的子模型}\only<3->{\sout{\textbf{特征}：人工设计的用于建模翻译的各种各样的子模型}}
\item \only<2>{\textbf{目标语语言模型}：大量单语数据训练的目标语建模模块}\only<3->{\sout{\textbf{目标语语言模型}：大量单语数据训练的目标语建模模块}}
\item \only<2>{...}\only<3->{\sout{...}}
\end{itemize}
\item<3-> 在\alert{神经机器翻译}时代，以上均不是必要的，而仅仅需要一个神经网络进行端到端建模
\end{itemize}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 编码器-解码器框架
\begin{frame}{编码器-解码器框架}
\begin{itemize}
\item 神经机器翻译遵循一种叫\alert{``编码器-解码器''}的结构
\begin{itemize}
\item<2-> \textbf{编码器}负责把源语言编码成一种句子表示形式(如：向量)
\item<3-> \textbf{解码器}负责利用这种表示逐词把目标语句子生成处理
\item<4-> 编码器和解码器之间的向量就是连接这两部分的句子表示
\end{itemize}
\end{itemize}
\vspace{0.5em}
\begin{tikzpicture}
\begin{scope}
\small{
\node [anchor=south west,minimum width=15em] (source) at (0,0) {\textbf{source}: 我\ \ \ \ 对\ \ \ \ 你\ \ \ \ 感到\ \ \ \ 满意};
\visible<3->{
\node [anchor=south west,minimum width=15em] (target) at ([yshift=12em]source.north west) {\textbf{target}: I\ \ am\ \ \ satisfied\ \ \ with\ \ \ you};
}
\visible<2->{
\node [anchor=center,minimum width=9.6em,minimum height=1.8em,draw,rounded corners=0.3em] (hidden) at ([yshift=6em]source.north) {};
\node [anchor=west,minimum width=1.5em,minimum size=1.5em,fill=ugreen!20] (cell01) at ([xshift=0.2em]hidden.west) {\footnotesize{.2}};
\node [anchor=west,minimum width=1.5em,minimum size=1.5em,fill=ugreen!10] (cell02) at (cell01.east) {\footnotesize{-1}};
\node [anchor=west,minimum width=1.5em,minimum size=1.5em,fill=ugreen!70] (cell03) at (cell02.east) {\footnotesize{6}};
\node [anchor=west,minimum width=1.5em,minimum size=1.5em,fill=ugreen!50] (cell04) at (cell03.east) {\footnotesize{5}};
\node [anchor=west,minimum width=1.5em,minimum size=1.5em,fill=ugreen!30] (cell05) at (cell04.east) {\footnotesize{.7}};
\node [anchor=west,minimum width=1.5em,minimum size=1.5em,fill=ugreen!10] (cell06) at (cell05.east) {\footnotesize{-2}};
}
\visible<2->{
\filldraw [fill=red!20,draw=white] (source.north west) -- (source.north east) -- ([xshift=-0.2em,yshift=-0.1em]hidden.south east) -- ([xshift=0.2em,yshift=-0.1em]hidden.south west);
}
\visible<3->{
\filldraw [fill=blue!20,draw=white] (target.south west) -- (target.south east) -- ([xshift=-0.2em,yshift=0.1em]hidden.north east) -- ([xshift=0.2em,yshift=0.1em]hidden.north west);
}
\visible<2->{
\draw [->,thick] (source.north west) -- ([xshift=0.2em,yshift=-0.1em]hidden.south west);
\draw [->,thick] (source.north east) -- ([xshift=-0.2em,yshift=-0.1em]hidden.south east);
}
\visible<3->{
\draw [->,thick] ([xshift=0.2em,yshift=0.1em]hidden.north west) -- (target.south west);
\draw [->,thick] ([xshift=-0.2em,yshift=0.1em]hidden.north east) -- (target.south east);
}
}
\visible<2->{
\node [anchor=south] (enclabel) at ([yshift=2em]source.north) {\large{\textbf{Encoder}}};
}
\visible<3->{
\node [anchor=north] (declabel) at ([yshift=-2em]target.south) {\large{\textbf{Decoder}}};
}
\visible<4->{
\node [anchor=west] (representation) at ([xshift=3.5em,yshift=3em]hidden.east) {\footnotesize{\textbf{句子的表示:}}};
\node [anchor=north west] (rline1) at (representation.south west) {\footnotesize{源语言句子被表示成一个实数}};
\node [anchor=north west] (rline2) at ([yshift=0.2em]rline1.south west) {\footnotesize{向量}\footnotesize{$(0.2,-1,6,5,0.7,-2)$}};
\node [anchor=north west] (rline3) at ([yshift=0.2em]rline2.south west) {\footnotesize{不要问0.2是什么意思，因为}};
\node [anchor=north west] (rline4) at ([yshift=0.2em]rline3.south west) {\footnotesize{只有系统自己才知道 :)}};
\begin{pgfonlayer}{background}
\visible<4->{
\node [rectangle,inner sep=0.4em,rounded corners=1pt,fill=ugreen!10,drop shadow] [fit = (representation) (rline1) (rline2) (rline3) (rline4)] (rlabel) {};
}
\end{pgfonlayer}
\draw[->,very thick,dashed] (hidden.east) .. controls +(east:1) and +(west:1) .. ([yshift=-1em]rlabel.north west);
}
\end{scope}
\end{tikzpicture}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
%%% 表示模型
\begin{frame}{基于连续空间表示模型的方法}
\begin{itemize}
\item 编码器-解码器框架的革命在于：\alert{它把传统基于符号的离散型知识转化为基于表示的连续型知识}
\begin{itemize}
\item 比如，对于一个短语，它可以对应一个文法规则的使用过程，而规则都是由离散的符号表示
\item 它也可以被表示为一种更加抽象的形式，通过向量记录短语的各个``属性''
\end{itemize}
\item<2-> 这种表示形式上的创新，让我们不再依赖离散符号系统的各种分解、组合，而直接在连续空间下学习这种表示
\begin{itemize}
\item 所谓``端到端''只是这种表示模型下的一种自然的学习范式
\end{itemize}
\end{itemize}
\visible<3->{
\begin{center}
\begin{tikzpicture}
\begin{scope}
\node [anchor=south west,draw,thick,red,minimum width=0.9in,minimum height=0.7in] (space1) at (0,0) {};
\node [anchor=south west,fill=blue,minimum width=0.1in,minimum height=0.1in] (unit1) at (0.2,0.8) {};
\node [anchor=south west,fill=ugreen,minimum width=0.1in,minimum height=0.1in] (unit2) at (0.7,0.3) {};
\node [anchor=south west,fill=blue,minimum width=0.1in,minimum height=0.1in] (unit3) at (1.3,1.3) {};
\node [anchor=south west,fill=ugreen,minimum width=0.1in,minimum height=0.1in] (unit4) at ([xshift=0.1em]unit3.south east) {};
\begin{pgfonlayer}{background}
\node [rectangle,inner sep=2pt,rounded corners=1pt,draw,thick] [fit = (unit3) (unit4)] (unitbox) {};
\end{pgfonlayer}
\draw [->] ([yshift=1pt]unit1.north) .. controls +(north:0.23) and +(west:0.2) .. ([yshift=0.2em,xshift=-1pt]unitbox.west);
\draw [->] ([xshift=1pt]unit2.east) .. controls +(east:0.5) and +(south:0.2) .. ([xshift=0.2em,yshift=-1pt]unitbox.south);
\node [anchor=south] (spacelabel1) at (space1.north) {\scriptsize{离散空间模型}};
\node [anchor=north] (captain1) at ([yshift=-0.5em]space1.south) {\scriptsize{(a) \textbf{统计机器翻译}}};
\end{scope}
\begin{scope}[xshift=1.3in]
\node [anchor=south west,draw,thick,red,minimum width=0.9in,minimum height=0.7in] (space1) at (0,0) {};
\node [anchor=south west,fill=blue,minimum width=0.1in,minimum height=0.1in] (unit1) at (0.2,0.8) {};
\node [anchor=south west,fill=ugreen,minimum width=0.1in,minimum height=0.1in] (unit2) at (0.7,0.3) {};
\node [anchor=south west,draw,thick,red,minimum width=0.9in,minimum height=0.7in] (space2) at (1.1in,0) {};
\node [anchor=south west,circle,fill=orange,minimum width=0.1in,minimum height=0.1in] (unit3) at (1.5in,1.3) {};
\draw [->] ([yshift=1pt]unit1.north) .. controls +(north:0.4) and +(west:2) .. ([yshift=0.0em,xshift=-1pt]unit3.west);
\draw [->] ([xshift=1pt]unit2.east) .. controls +(east:1.5) and +(south:1) .. ([xshift=0.0em,yshift=-1pt]unit3.south);
\node [anchor=south] (spacelabel1) at (space1.north) {\scriptsize{离散空间模型}};
\node [anchor=south] (spacelabel2) at (space2.north) {\scriptsize{连续空间模型}};
\node [anchor=north] (captain1) at ([yshift=-0.5em,xshift=1em]space1.south east) {\scriptsize{(b) \textbf{神经机器翻译}}};
\end{scope}
\end{tikzpicture}
\end{center}
}
\end{frame}
%%%------------------------------------------------------------------------------------------------------------
\section{循环神经网络翻译模型及注意力机制}
%%%------------------------------------------------------------------------------------------------------------
......@@ -137,5 +387,8 @@
%%%------------------------------------------------------------------------------------------------------------
\subsection{深层网络}
%%%------------------------------------------------------------------------------------------------------------
\section{其它应用}
\end{CJK}
\end{document}
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