A Lightweight Designed Beamer Template of Weekly Survey

   blog 13 minutes read (About 1948 words) 0 visits

Introduction

Nowadays, I’m working on a weekly report for my research group. Finding a concise and academic slides template is a need for us to represent our finds and ideas. Based on the PKU_beamer_lightweight_designed, I adjust some details according to my preference and share the tutorial in this post.

Adjusted Details

This a lightweight designed beamer template of weekly survey with the version of Xidian University, which is useful for reporting your research work academically and concisely.

  • Theme Color
    I use the official red color of Xidian University as the theme color. The hex code of the color is #B0252A.
  • Main Font
    I prefer the fontstyle of Times New Roman so I need to additionally include fontspec package.
    1
    2
    \usepackage{fontspec}
    \setsansfont{Times New Roman}
  • Remove Specific Frame From the Healine
    In beamer, the default frames will be counted in the headline which is sometimes not suitable. Such as the title page, overline page and so on are usually independent pages.

To remove them from the headline, I use the following code.

configuration
1
2
3
4
5
6
7
8
9
10
11
12
\makeatletter
\let\beamer@writeslidentry@miniframeson=\beamer@writeslidentry%
\def\beamer@writeslidentry@miniframesoff{%
  \expandafter\beamer@ifempty\expandafter{\beamer@framestartpage}{}% does not happen normally
  {%else
    % removed \addtocontents commands
    \clearpage\beamer@notesactions%
  }
}
\newcommand*{\miniframeson}{\let\beamer@writeslidentry=\beamer@writeslidentry@miniframeson}
\newcommand*{\miniframesoff}{\let\beamer@writeslidentry=\beamer@writeslidentry@miniframesoff}
\makeatother

Before those frames, we just turn off the frames style but if we next to count frame we also need to open this by the command of \miniframeson. And the section*{} ensures the frame removes the number in some section.

usage
1
2
3
4
5
6
7
\miniframesoff
\begin{frame}
    \section*{}
    \begin{center}
        {\Huge \textit{Thanks for you3fr listening!}}
    \end{center}
\end{frame}

XidianU.sty Codes

XidianU.sty
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
\mode<presentation>

\newif\ifbeamer@secheader
\beamer@secheaderfalse

\ProcessOptionsBeamer

\useoutertheme[subsection=false]{smoothbars}
\makeatletter
\newcommand{\frameofframes}{/}
\newcommand{\setframeofframes}[1]{\renewcommand{\frameofframes}{#1}}
\setbeamertemplate{footline} 
  {%
    \begin{beamercolorbox}[colsep=1.5pt]{upper separation line foot}
    \end{beamercolorbox}
    \begin{beamercolorbox}[ht=2.5ex,dp=1ex,%
      leftskip=.3cm,rightskip=.3cm plus1fil]{author in head/foot}%
      {\usebeamerfont{author in head/foot}\insertshortauthor}%
      \hfill%
      {\usebeamerfont{title in head/foot}\insertshorttitle}%
      \hfill%
      {\usebeamerfont{frame number}\usebeamercolor[fg]{frame number}\insertframenumber~\frameofframes~\inserttotalframenumber}
    \end{beamercolorbox}%
    \begin{beamercolorbox}[colsep=1.5pt]{lower separation line foot}
    \end{beamercolorbox}
  }
\makeatother

\useinnertheme{circles}
%\useinnertheme{rectangles}
%\useoutertheme{default}
%\useinnertheme[shadow=true]{rounded}

\definecolor{xidian}{HTML}{B0252A}
% \xdefinecolor{xidian}{cmyk}{0,1,1,0.45}%{rgb}{0.543,0.0,0.0703} %{cmyk}{0,100,100,45}%{rgb}{0.5,0.0,0.0}  %RGB#820010
\xdefinecolor{xidian_gold}{cmyk}{0,0.35,0.75,0.05}
\xdefinecolor{xidian_blue}{cmyk}{0.6,0.35,0.0,0.4}
\xdefinecolor{xidian_darkblue}{cmyk}{1.0,0.6,0.0,0.5}
\xdefinecolor{xidian_gray}{cmyk}{0.0,0.0,0.08,0.55}
\xdefinecolor{xidian_dirt}{cmyk}{0.0,0.2,0.35,0.3}
\xdefinecolor{xidian_orange}{cmyk}{0.0,0.7,1.0,0.0}
\xdefinecolor{xidian_green}{cmyk}{0.2,0.0,1.0,0.15}
\xdefinecolor{xidian_darkgreen}{cmyk}{0.6,0.5,1.0,0.45}
\xdefinecolor{pantone_gold}{RGB}{135,103,79}
\xdefinecolor{pantone_silver}{RGB}{138,141,143}
\xdefinecolor{WM_Gold}{cmyk}{0.09,0.29,0.66,0.24}
\setbeamercolor{footline}{bg=xidian}
%\setbeamercolor{frametitle}{bg=white!70!pantone_gold,fg=xidian}
\setbeamercolor{frametitle}{bg=white,fg=xidian}
\setbeamercolor{title}{bg=xidian}
%\setbeamerfont{frametitle}{size=\large}
\setbeamerfont{frametitle}{series=\bfseries,size=\large}%,parent=structure}
\setbeamerfont{footline}{series=\bfseries}

\setbeamertemplate{navigation symbols}{}
\setbeamertemplate{bibliography item}[text]
\setbeamertemplate{caption}[numbered]

\beamertemplateshadingbackground{white!5}{white}

\setbeamercolor{palette primary}{use=structure,fg=white,bg=structure.fg}
\setbeamercolor{palette secondary}{use=structure,fg=white,bg=structure.fg!95!black}%{use=structure,fg=white,bg=structure.fg!90!black}
\setbeamercolor{palette tertiary}{use=structure,fg=white,bg=structure.fg!90!black}
\setbeamercolor{palette quaternary}{fg=white,bg=structure.fg!85!black}
%\setbeamercolor*{sidebar}{use=structure,bg=structure.fg}
\setbeamercolor{titlelike}{parent=palette primary}

%% try
\setbeamercolor{block title}{bg=xidian_blue,fg=white}
\setbeamercolor{block body}{bg=xidian_blue!10}

\BeforeBeginEnvironment{definition}{%
    \setbeamercolor{block title}{bg=xidian_blue,fg=white}
    \setbeamercolor{block body}{bg=xidian_blue!10}
}
\AfterEndEnvironment{definition}{
 \setbeamercolor{block title}{bg=xidian_blue,fg=white}
 \setbeamercolor{block body}{bg=xidian_blue!10}
}

\BeforeBeginEnvironment{theorem}{%
    \setbeamercolor{block title}{bg=xidian_orange,fg=white}
    \setbeamercolor{block body}{bg=xidian_orange!10}
}
\AfterEndEnvironment{theorem}{
 \setbeamercolor{block title}{bg=xidian_blue,fg=white}
 \setbeamercolor{block body}{bg=xidian_blue!10}
}

\BeforeBeginEnvironment{proposition}{%
    \setbeamercolor{block title}{bg=xidian_orange,fg=white}
    \setbeamercolor{block body}{bg=xidian_orange!10}
}
\AfterEndEnvironment{proposition}{
 \setbeamercolor{block title}{bg=xidian_blue,fg=white}
 \setbeamercolor{block body}{bg=xidian_blue!10}
}

\setbeamercolor*{block title example}{use={normal text,example text},bg=white!70!pantone_gold,fg=xidian}
\setbeamercolor{fine separation line}{}
\setbeamercolor{item projected}{fg=white}
\setbeamercolor{palette sidebar primary}{use=normal text,fg=normal text.fg}
\setbeamercolor{palette sidebar quaternary}{use=structure,fg=structure.fg}
\setbeamercolor{palette sidebar secondary}{use=structure,fg=structure.fg}
\setbeamercolor{palette sidebar tertiary}{use=normal text,fg=normal text.fg}
%\setbeamercolor{palette sidebar quaternary}{fg=white}
\setbeamercolor{section in sidebar}{fg=brown}
\setbeamercolor{section in sidebar shaded}{fg=grey}
\setbeamercolor{separation line}{}
\setbeamercolor{sidebar}{bg=xidian}
\setbeamercolor{sidebar}{parent=palette primary}
\setbeamercolor{structure}{fg=xidian}
\setbeamercolor{subsection in sidebar}{fg=brown}
\setbeamercolor{subsection in sidebar shaded}{fg=grey}

\AtBeginSection[]{
	\begin{frame}
		\tableofcontents[sectionstyle=show/shaded,subsectionstyle=hide,subsubsectionstyle=hide]
	\end{frame}
}  


\setbeamercolor{postgreen}{fg=black,bg=example text.fg!75!black!10!bg}
\setbeamercolor{postred}{fg=black,bg=white!70!pantone_gold}
\setbeamercolor{postblue}{fg=black,bg=xidian_blue!10}
%\AtBeginSubsection[]{
%	\begin{frame}
%		\tableofcontents[sectionstyle=show/shaded,subsectionstyle=hide,subsubsectionstyle=hide]
%	\end{frame}
%}

\mode
<all>

main.tex

main.tex
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
\documentclass[10pt,hyperref={colorlinks,citecolor=blue,urlcolor=xidian_blue,linkcolor=}]{beamer}
\usepackage{XidianU}
\usepackage{fontspec}
\setsansfont{Times New Roman}

\usepackage{lipsum}
%\usepackage[scheme = plain]{ctex}
\usepackage{charter} % Nicer fonts
% other packages
\usepackage{latexsym,amsmath,xcolor,multicol,booktabs,calligra}
\usepackage{amssymb}
\usepackage{graphicx}
\usepackage{subcaption}
\usepackage{bm}
\usepackage{natbib}
\usepackage{wrapfig}
\usepackage{amsfonts} 
\usepackage{ragged2e}
\usepackage{parskip}

\apptocmd{\frame}{}{\justifying}{} % Allow optional arguments after frame.
\newcommand{\theHalgorithm}{\arabic{algorithm}}
\theoremstyle{plain}
\newtheorem{axiom}{Axiom}
\newtheorem{claim}[axiom]{Claim}
\newtheorem{assumption}{Assumption}
\newtheorem{remark}{Remark}
\newtheorem{proposition}{Proposition}
\setbeamertemplate{theorems}[numbered]

% change for your title page information
\author[Zhihao Li]{Zhihao Li}
\title{Research Survey 1}
\subtitle{Why Is Prompt Tuning for Vision-Language Models Robust to Noisy Labels?}
\institute{School of Computer Science and Technology\\Xidian University}
\date{February 1, 2024}

% official colors match with the Xidian color
\def\cmd#1{\texttt{\color{red}\footnotesize $\backslash$#1}}
\def\env#1{\texttt{\color{blue}\footnotesize #1}}
\definecolor{deepblue}{rgb}{0,0,0.5}
\definecolor{deepred}{rgb}{0.6,0,0}
\definecolor{deepgreen}{rgb}{0,0.5,0}
\definecolor{halfgray}{gray}{0.55}

\show\hss

\makeatletter
\let\beamer@writeslidentry@miniframeson=\beamer@writeslidentry%
\def\beamer@writeslidentry@miniframesoff{%
  \expandafter\beamer@ifempty\expandafter{\beamer@framestartpage}{}% does not happen normally
  {%else
    % removed \addtocontents commands
    \clearpage\beamer@notesactions%
  }
}
\newcommand*{\miniframeson}{\let\beamer@writeslidentry=\beamer@writeslidentry@miniframeson}
\newcommand*{\miniframesoff}{\let\beamer@writeslidentry=\beamer@writeslidentry@miniframesoff}
\makeatother

\begin{document}
{
\begin{frame}
    \titlepage
    \begin{figure}[htpb]
        \begin{center}
            \includegraphics[width=0.2\linewidth]{Figures/XDUlogo.jpg}
        \end{center}
    \end{figure}
\end{frame}
}
\section{Summary}
\begin{frame}{Weekly Work}
\begin{enumerate}
    \item Read the paper of \textit{Why Is Prompt Tuning for Vision-Language Models Robust to Noisy Labels?};
    \item Learn about some concepts;
\end{enumerate}
\end{frame}

\begin{frame}{A prompt tuning process is highly
robust to label noises.}


\begin{enumerate}
    \item \textbf{Interest}: Studying the key reasons contributing to the robustness of the prompt tuning.
paradigm.
    \item \textbf{Findings}: \begin{enumerate}
        \item the fixed classname tokens provide a strong regularization to the optimization of the model, reducing gradients induced by the noisy samples;
        \item the powerful pre-trained image-text embedding that is learned from diverse and generic web data provides strong prior knowledge for image classification.
    \end{enumerate}
\end{enumerate}
\end{frame}

\begin{frame}{Author's Contributions}
    \begin{itemize}
        \item We demonstrate that \textbf{prompt tuning for pre-trained vision-language models (e.g., CLIP) is more robust to noisy labels} than traditional transfer learning approaches, such as model fine-tuning and linear probes.
        \item We further demonstrate that \textbf{prompt tuning robustness can be further enhanced through the use of a robust training objective.}
        \item We conduct an extensive analysis on why prompt tuning is robust to noisy labels to \textbf{discover which components contribute the most to its robustness.}
        \item We \textbf{propose a simple yet effective method for unsupervised prompt tuning}, showing that randomly selected noisy pseudo labels can be effectively used to enhance CLIP zero-shot performance. The proposed robust prompt tuning outperformed prior work on a variety of datasets, even though noisier pseudo-labels are used for self-training.
    \end{itemize}
\end{frame}

\section{Motivations}
\begin{frame}{Mathematical Models}
    \begin{itemize}
        \item CLIP\\
        In the case of image classification, a normalized image embedding $\boldsymbol{f}^{\:v}$ is obtained by passing an image through CLIP's visual encoder, and a set of normalized class embeddings $[\boldsymbol{f_i^{\:t}}]_{i=1}^K$ by feeding template prompts of the form "A photo of a" into CLIP's text encoder.
        \begin{equation}
        Pr(y=i|\boldsymbol{x})=\frac{\exp(sim(\boldsymbol{f}^{\:v},\boldsymbol{f}^{\:t}_i))/\tau}{\sum_{j=1}^K\exp(sim(\boldsymbol{f}^{\:v},\boldsymbol{f}^{\:t}_j))/\tau}
        \end{equation}
        \item Prompt Tuning\\
        The name of a class c is first converted into a classname embedding $\boldsymbol{w}\in R^d$ and prepended with a sequence of $M$ learnable tokens $\boldsymbol{p_m}\in R^d$ shared across all classes.
        \begin{equation}
        P_c=[\boldsymbol{p_1}, \boldsymbol{p_2}, \cdots, \boldsymbol{p_M}, \boldsymbol{w_c}]\rightarrow \boldsymbol{f}^{\:t}_c
        \end{equation}
        CoOp optimizes the shared learnable tokens $\boldsymbol{p_1}, \boldsymbol{p_1}, \cdots, \boldsymbol{p_M}$ on a small labeled dataset $D = [(\boldsymbol{x_i}, c_i)^N_{i=1}]$ to minimize the cross-entropy loss:
        \begin{equation}
            L_{CE}=-E_{(\boldsymbol{x},c)\in D}[\log Pr(y=c|\boldsymbol{x})]
        \end{equation}
    \end{itemize}
\end{frame}

\begin{frame}{Mathematical Models}
    \begin{itemize}
        \item Robust Prompt Tuning\\
        Further enhance this robustness by optimizing the learnable prompts using the generalized cross-entropy (GCE) loss:
        \begin{equation}
            L_{GCE}=E_{(\boldsymbol{x},c)\in D}[\frac{1-Pr(y=c|\boldsymbol{x})^q}{q}]
        \end{equation}
        \item Author's Conclusion: $q = 0.7$ leads to overall good performance across several experimental settings.
    \end{itemize}
\end{frame}

\section{Robustness Analysis}

\begin{frame}{Pre-trained CLIP Generates Effective Class Embeddings}
    \vspace{-1em}
    \begin{figure}
    \includegraphics[width=0.9\textwidth]{Figures/Survey1/models.png}
    \label{fig: Models}
    \end{figure}\vspace{-0.7em}
    \begin{itemize}
        \item Classifier-R v.s. Classifier-C: CLIP class embeddings provide a strong initialization for few-shot learning.\vspace{-0.5em}
        \item TEnc-FT v.s. Classifier-C: The highly expressive CLIP text encoder can easily overfit to the noisy labels.\vspace{-0.5em}
        \item Prompt Tuning v.s. Classifiers: The text encoder is essential for providing a strong but informative regularization of the text embeddings to combat noisy inputs.\vspace{-0.5em}
        \item Prompt Tuning v.s. TEnc-FT: The text encoder should be fixed to prevent overfitting.
    \end{itemize}
\end{frame}
\begin{frame}{Other Aspects of Robustness}
\begin{itemize}
    \item \textbf{Effectiveness of Prompt}
    \item \textbf{Prompt Tuning Suppresses Noisy Gradients}
    \item \textbf{Generalization Across Model Architectures}
    \item \textbf{Robustness to Correlated Label Noise}
\end{itemize}
\end{frame}

\section{Robust UPL}
\begin{frame}{Improve UPL in Unsupervised Prompt Tuning}
    \vspace{-1em}
    \begin{figure}
    \includegraphics[width=\textwidth]{Figures/Survey1/UPL.png}
    \label{fig: UPL}
    \end{figure}\vspace{-0.8em}
\begin{itemize}
    \item Baseline UPL\begin{itemize}
        \item Phase 1: Leverage pre-trained CLIP to generate pseudo labels for unlabeled images.
        \item Phase 2: Select \textbf{the $K$ most confident samples per class} to optimize the learnable tokens through the typical prompt-tuning optimization process (described in CoOp).
        \end{itemize}
    \item Robust UPL\\
    Based on UPL, \textbf{randomly sample $K$ training samples} and optimize the prompt with the \textbf{robust GCE loss}.
\end{itemize}
\end{frame}
\section{Next Stage}
\begin{frame}{New Plans for Next Week}
\begin{enumerate}
    \item Reproduce the most of results about this paper.
    \item Survey other relavent methods in this domain.
\end{enumerate} 
\end{frame}

\miniframesoff
\begin{frame}
    \section*{}
    \begin{center}
        {\Huge \textit{Thanks for you3fr listening!}}
    \end{center}
\end{frame}

\end{document}

Template Overview

Contributors

References

A Lightweight Designed Beamer Template of Weekly Survey

https://lzhms.github.io/blog/WeeklyReport/

Author

Zhihao Li

Posted on

2024-02-18

Updated on

2024-12-24

Licensed under

 

Comments

  Follow

Catalogue

Categories

Tags

Embedded System3
Project Tools4
Hexo1
Computer Graphics2
Professional Knowledge1
Diffusion Model1
AIGC1
Project Habits2
GAN1
Mathematical Modeling1
Data Visualization1
OpenSSL1
Research Habits2
Paper Writing1
Linux3
View All Tags >>
×