我在使用 LaTeX 制作表格时遇到了大问题。我需要您的帮助，使下表更短、更紧凑。到目前为止，我得到的结果并不令人满意。前两列和最后一列可以合并为一列。所以，保留 3 列。句子之间的间距可以减少，如果可能的话，只在 2 页上。请帮忙。
代码：
%% ----------------------------------------------------------------
%% Thesis.tex -- MAIN FILE (the one that you compile with LaTeX)
%% ----------------------------------------------------------------
% Set up the document
\documentclass[a4paper, 11pt, arial, oneside]{Thesis} % Use the "Thesis" document class, based on the ECS Thesis style by Steve Gunn
\usepackage{charter}
\usepackage[scaled=0.92]{helvet}
%\usepackage [T1] {fontenc}
%\renewcommand\familydefault{phv}
%\graphicspath{{figs/}} % Location of the graphics files (set up for graphics to be in PDF format)
% Include any extra LaTeX packages required
\usepackage[square, numbers, comma, sort&compress]{natbib} % Use the "Natbib" style for the references in the Bibliography
\usepackage{verbatim} % Needed for the "comment" environment to make LaTeX comments
\usepackage{vector} % Allows "\bvec{}" and "\buvec{}" for "blackboard" style bold vectors in maths
%\hypersetup{urlcolor=blue, colorlinks=true} % Colours hyperlinks in blue, but this can be distracting if there are many links.
\usepackage[toc,page]{appendix}
\usepackage{graphicx}
\usepackage{romannum}
\usepackage{enumerate}
%\usepackage{booktabs}
%\usepackage[hmargin={30mm,25mm},vmargin=25mm]{geometry}
%\usepackage{array,ragged2e}
%\newcolumntype{R}[1]{>{\RaggedRight}p{#1}}
%\usepackage{enumitem}
%\usepackage{etoolbox}
%\AtBeginEnvironment{table}{%
% \setlist[itemize]{ nosep, % <-- new list setup
% leftmargin = *,
% before = \vspace{-0.6\baselineskip},
% after = \vspace{-\baselineskip}
% }
%}
%\usepackage{showframe}
%\usepackage{geometry}
\usepackage{ragged2e, tabularx}
\usepackage{array, booktabs, longtable, makecell, threeparttablex}
\newcolumntype{R}[1]{>{\RaggedRight\hspace{0pt}}p{#1}}
\usepackage[skip=1ex,labelfont=bf,font=small]{caption}
\usepackage[shortlabels]{enumitem}
\usepackage{etoolbox}
\AtBeginEnvironment{longtable}{%
\small % for better fit text into cells
\setlength{\LTcapwidth}{\linewidth} % that caption width is equal table width
\setlist[itemize]{ nosep, % <-- new list setup
leftmargin = *,
before = \vspace{-\baselineskip},
after = \vspace{-\baselineskip}
}
}% end of AtBeginEnvironment
\AtBeginEnvironment{table}{%
\setlist[itemize]{nosep,
wide,%leftmargin = *,
before = \vspace{-\baselineskip},
after = \vspace{-\baselineskip}
}
}% end of AtBeginEnvironment
\newcolumntype{L}{>{\RaggedRight\setlength\parskip{0.2\baselineskip}\arraybackslash}X}
\setlength\extrarowheight{1pt}
\usepackage{siunitx}
\usepackage[space]{grffile}
\usepackage{latexsym}
\usepackage{textcomp}
\usepackage{longtable}
\usepackage{pdflscape}
\usepackage{multirow,booktabs}
\usepackage{amsfonts,amsmath,amssymb}
\usepackage{url}
\usepackage{hyperref}
\hypersetup{colorlinks=true,pdfborder={0 0 0},linkcolor=blue,urlcolor=blue,citecolor=red}
% You can conditionalize code for latexml or normal latex using this.
\newif\iflatexml\latexmlfalse
\providecommand{\tightlist}{\setlength{\itemsep}{0pt}\setlength{\parskip}{0pt}}%
\DeclareGraphicsExtensions{.pdf,.PDF,.png,.PNG,.jpg,.JPG,.jpeg,.JPEG}
%\usepackage[utf8]{inputenc}
\usepackage[british]{babel}
% correct bad hyphenation here
\hyphenation{op-tical net-works semi-conduc-tor}
\usepackage{siunitx} % For units
\newcommand\mmfeed[1]{\textcolor{red}{#1}}
\usepackage{xcolor}
\usepackage{caption}
\usepackage{subcaption}
\usepackage{lipsum}
\captionsetup[figure]{labelfont={bf,small},textfont={small}}
\captionsetup[subfloat]{labelfont={bf,small}, textfont={small}, subrefformat=parens} %<-----designing subcaption
\newcommand{\myfigref}[2]{~\ref{#1}.\subref{#2}}% <---- a new macro for referring to a subfigure
%
%\newpage
%\bibliographystyle{plain}
%\bibliography{bibliografia}
% Change the text font
\renewcommand{\rmdefault}{phv} % Arial
\renewcommand{\sfdefault}{phv} % Arial
%
\usepackage{multirow}
\usepackage{cleveref}
% I added many usepackages
%% ----------------------------------------------------------------
\begin{document}
\begin{longtable}{
R{\dimexpr0.15\linewidth-2\tabcolsep}
R{\dimexpr0.25\linewidth-2\tabcolsep}
R{\dimexpr0.25\linewidth-2\tabcolsep}
R{\dimexpr0.25\linewidth-2\tabcolsep}
R{\dimexpr0.15\linewidth-4\tabcolsep}
}
\caption{The state-of-the-art in taxonomy, detection, extraction and pattern recognition in urban applications}
\label{T2.5} \\
\toprule
%& \multicolumn{3}{c}{Consequences of the expansion of cities}
%\\
%\cmidrule(lr){2-4}
\textbf{Application} & \textbf{Taxonomy/quantification algorithms} & \textbf{Merits} & \textbf{Limitations} & \textbf{Example}
\\
\midrule
\endfirsthead
\caption{The state-of-the-art in \dots\hfill (cont. from previous page)}
%\\
%\toprule
%& \multicolumn{3}{c}{Consequences of the expansion of cities}
%\\
%\cmidrule(lr){2-4}
%& Temperature & Air quality & Water quality
\\
\midrule
\endhead
\multicolumn{4}{r}{\footnotesize\textit{continued on the next page}}
\endfoot
\endlastfoot
%%%%
Seismic building structural types (SBSTs)
& \begin{itemize}
\item Support vector machine (SVM) and random forest (RF)
\end{itemize}
& \begin{itemize}
\item Classify the combination of different remote sensing data
\item Derive sets of valuable features to characterise the urban environment
\item Model an effective earthquake loss and spatial distribution
\item ABTSVM classifier outperforms other multi-class SVM classifiers
\end{itemize}
& \begin{itemize}
\item Hierarchical supervised classification scheme has uncertainties in separating SBSTs
\item Performance depends on the ranked features
\item Accuracy is subject to the addition of further features and subset based categories
\end{itemize}
& \begin{itemize}
\item \cite{geiss2015}
\end{itemize} \\
\midrule
Urban change detection: landuse transitions
& \begin{itemize}
\item Scene classification with a bag-of-visual-words (BOVW)
\end{itemize}
& \begin{itemize}
\item Obtain semantic scene classes
\item Effectively analyse landuse changes
\item Satisfactory accuracy
\end{itemize}
& \begin{itemize}
\item Time-consuming
\item Very difficult to achieve the direct selection of the “from-to” samples from the dataset
\item Classification performance in certain cases is negatively affected by the redundancy of information
\end{itemize}
& \begin{itemize}
\item \cite{wu2016}
\end{itemize} \\
\midrule
Urban landcover and landuse classification
& \begin{itemize}
\item Binary tree SVM based on Jeffries–Matusita (JM) distance
\end{itemize}
& \begin{itemize}
\item Improve classification accuracy
\item Classify hyperspectral images adequately
\item Ease of interpretation of urban classes
\end{itemize}
& \begin{itemize}
\item Confusion between road and bare soil classes
\item Instability and complexity in the structure and parameters of the binary tree SVM
\end{itemize}
& \begin{itemize}
\item \cite{du2012}
\end{itemize} \\
\midrule
Monitoring changes in impervious surfaces
& \begin{itemize}
\item MRGU
\end{itemize}
& \begin{itemize}
\item Integrate multiple endmember spectral mixture analysis (MESMA), analysis of regression residuals, spatial statistics (Getis Ord) with Moran’s I, and urban growth theories in an effective manner
\item Quantify and identify the magnitude of impervious surface changes and their spatial distribution
\end{itemize}
& \begin{itemize}
\item Not universally applicable due to specific behaviour of maximum noise fraction (MNF)
\item Difficult to use for quantifying changes in urban centres
\item Performance (regression residuals) is subject to the structure of urban regions
\end{itemize}
& \begin{itemize}
\item \cite{shahtahmassebi2016}
\end{itemize} \\
\midrule
Change detection
& \begin{itemize}
\item Unsupervised Neural network and feature transformation
\end{itemize}
& \begin{itemize}
\item Deep architecture (SDAE) for better representation of the relationships between feature- and pixel-pair
\item Mapping based FCA and function learning to highlight change in a robust manner
\item Denoising autoencoder
\end{itemize}
& \begin{itemize}
\item Uncertainty in the feature-pair
\item High computational cost
\item Complexity structures of stacked denoising autoencoders (SDAE)
\item Many constraints are required to extract useful features
\end{itemize}
& \begin{itemize}
\item \cite{zhang2016}
\end{itemize} \\
\midrule
Water quality, and sustainable water resources management
& \begin{itemize}
\item Integrated data fusion and mining (IDFM) and artificial neural network (ANN)
\end{itemize}
& \begin{itemize}
\item A near real-time monitoring and the early warning system
\item Efficiency
\item Forecasting reliability
\item Potential for local adoption
\end{itemize}
& \begin{itemize}
\item Prediction accuracy may be effected by uncertainties in the fused data
\item A large number of variables is required to overcome the uncertainty
\item Not applicable for regional meteorology parameters
\end{itemize}
& \begin{itemize}
\item \cite{imen2015}
\end{itemize} \\
\midrule
Air temperature estimation
& \begin{itemize}
\item SVM
\end{itemize}
& \begin{itemize}
\item Fully automated method
\item SVM regression is robust
\item Regression errors can be modelled at the pixel level, improving accuracy estimation
\end{itemize}
& \begin{itemize}
\item Requires expert users to apply SVM
\item Does not work well under non cloud-free conditions and require in situ measurements
\item Regression error distribution is insufficient
\end{itemize}
& \begin{itemize}
\item \cite{moser2015}
\end{itemize} \\
\midrule
Fine-scale population estimation for urban management, emergency response and epidemiological
& \begin{itemize}
\item RF
\item Linear regression modelling
\end{itemize}
& \begin{itemize}
\item Able to classify building types and extract their footprints in the heterogeneous urban areas
\item Improved classification accuracy
\item Ease of adoption
\end{itemize}
& \begin{itemize}
\item Subject to the accuracy of the selected morphology filter
\item Use of large numbers of metrics and variables for building type classification
\item Building background metrics do not show its advantage in the block classification
\item Classification uncertainty for non-residential buildings
\end{itemize}
& \begin{itemize}
\item \cite{xie2015}
\end{itemize} \\
\midrule
Renewable energy and urban feature extraction
& \begin{itemize}
\item Shadow detection and building geometry identification
\end{itemize}
& \begin{itemize}
\item Easy to apply
\item Sufficient to generate 3D model of urban buildings
\item Reliable analysis of the solar energy potential
\item Identify the availability of 3D surfaces
\item Flexibility and feasibility
\end{itemize}
& \begin{itemize}
\item Not fully automated
\item Not suitable for dense urban areas
\item Sensitive to the quality of satellite images
\end{itemize}
& \begin{itemize}
\item \cite{kadhim2015a}
\end{itemize} \\
\midrule
Impervious surfaces estimation
& \begin{itemize}
\item SVM
\item RF
\end{itemize}
& \begin{itemize}
\item Increased classification accuracy
\item Does not depend on combinations of features
\item Data can be fused to optimise parameters efficiently
\item Ease of application
\end{itemize}
& \begin{itemize}
\item Needs many texture matrices
\item Inability to handle the confusion in shaded areas and bare soil
\item Over-fitting
\end{itemize}
& \begin{itemize}
\item \cite{zhang2015}
\end{itemize} \\
\midrule
\end{longtable}
\end{document}
答案1
不知道哪个您所采用的文档类别数量几乎是无限的Thesis，以下解决方案将使用report文档类别，并假设纸张大小为 A4，边距为 2.5 厘米。显然，您可以根据需要随意调整这些设置。
我的主要建议是将第 1、2 和 5 列变窄，相反，将第 3 和第 4 列变宽。这样，确实可以将其longtable放在两整页上。
itemize除非列表中确实有多个项目，否则我不会在第二列中使用环境。
\documentclass[a4paper,11pt,arial,oneside]{report}%{Thesis} % which "Thesis"?!
\usepackage[a4paper,margin=2.5cm]{geometry} % choose page parameters suitably
%\usepackage{charter} % no need to load this package, right?
\usepackage[scaled=0.92]{helvet}
\usepackage[T1]{fontenc}
\renewcommand\familydefault{\sfdefault} % use Helvetica as main text font
\usepackage[square, numbers, comma, sort&compress]{natbib}
\usepackage{verbatim,vector}
\usepackage[toc,page]{appendix}
\usepackage{graphicx,romannum}
\usepackage{ragged2e, tabularx}
\usepackage{array, booktabs, longtable, makecell, threeparttablex}
\newcolumntype{R}[1]{>{\RaggedRight\hspace{0pt}\arraybackslash}p{#1}}
\usepackage[skip=1ex,labelfont=bf,font=small]{caption}
\usepackage[shortlabels]{enumitem}
\usepackage{etoolbox}
\AtBeginEnvironment{longtable}{%
\small
\setlength{\LTcapwidth}{\linewidth}
\setlist[itemize]{ nosep, % <-- new list setup
leftmargin = *,
before = \vspace{-\baselineskip},
after = \vspace{-\baselineskip}
}
}
\AtBeginEnvironment{table}{%
\setlist[itemize]{nosep,
wide,%leftmargin = *,
before = \vspace{-\baselineskip},
after = \vspace{-\baselineskip}
}
}
\setlength\extrarowheight{1pt}
\usepackage{siunitx}
\usepackage[space]{grffile}
%%%%\usepackage{latexsym} % deprecated and superseded by "amssymb"
\usepackage{textcomp}
\usepackage{longtable}
\usepackage{pdflscape}
\usepackage{multirow,booktabs}
\usepackage{amsfonts,amsmath,amssymb}
% You can conditionalize code for latexml or normal latex using this.
\newif\iflatexml\latexmlfalse
\providecommand{\tightlist}{\setlength{\itemsep}{0pt}\setlength{\parskip}{0pt}}%
\usepackage[utf8]{inputenc}
\usepackage[british]{babel}
% correct bad hyphenation here
\hyphenation{op-tical net-works semi-conduc-tor quan-ti-fi-ca-tion}
\usepackage{siunitx} % For units
\newcommand\mmfeed[1]{\textcolor{red}{#1}}
\usepackage{xcolor}
\usepackage{caption}
\usepackage{subcaption}
\usepackage{lipsum}
\captionsetup[figure]{labelfont={bf,small},textfont={small},
skip=0.333\baselineskip}
\captionsetup[subfloat]{labelfont={bf,small}, textfont={small},
subrefformat=parens}
\newcommand{\myfigref}[2]{~\ref{#1}.\subref{#2}}
\usepackage{url}
\usepackage{hyperref} % load this package (almost) last
\hypersetup{colorlinks=true,
%pdfborder={0 0 0},
linkcolor=blue,urlcolor=blue,citecolor=red}
\usepackage{cleveref}
\begin{document}
\begin{longtable}{@{}
R{\dimexpr0.15\linewidth-1\tabcolsep}
R{\dimexpr0.15\linewidth-2\tabcolsep}
R{\dimexpr0.30\linewidth-2\tabcolsep}
R{\dimexpr0.30\linewidth-2\tabcolsep}
R{\dimexpr0.10\linewidth-1\tabcolsep}
@{}}
\caption{State-of-the-art in taxonomy, detection, extraction
and pattern recognition in urban applications}
\label{T2.5} \\
\toprule
\textbf{Application} & \textbf{Taxonomy\slash quantification algorithms} & \textbf{Merits} & \textbf{Limitations} & \textbf{Example} \\
\midrule
\endfirsthead
\caption{The state-of-the-art in \dots\hfill (cont.\ from previous page)} \\
\midrule
\endhead
\multicolumn{5}{r@{}}{\footnotesize\textit{continued on the next page}}
\endfoot
\bottomrule
\endlastfoot
%%%%
Seismic building structural types (SBSTs)
& Support vector machine (SVM) and random forest (RF)
& \begin{itemize}
\item Classify the combination of different remote sensing data
\item Derive sets of valuable features to characterise the urban environment
\item Model an effective earthquake loss and spatial distribution
\item ABTSVM classifier outperforms other multi-class SVM classifiers
\end{itemize}
& \begin{itemize}
\item Hierarchical supervised classification scheme has uncertainties in separating SBSTs
\item Performance depends on the ranked features
\item Accuracy is subject to the addition of further features and subset based categories
\end{itemize}
& \begin{itemize}
\item \cite{geiss2015}
\end{itemize} \\
\midrule
Urban change detection: landuse transitions
& Scene classification with a bag-of-visual-words (BOVW)
& \begin{itemize}
\item Obtain semantic scene classes
\item Effectively analyse landuse changes
\item Satisfactory accuracy
\end{itemize}
& \begin{itemize}
\item Time-consuming
\item Very difficult to achieve the direct selection of the “from-to” samples from the dataset
\item Classification performance in certain cases is negatively affected by the redundancy of information
\end{itemize}
& \begin{itemize}
\item \cite{wu2016}
\end{itemize} \\
\midrule
Urban landcover and landuse classification
& Binary tree SVM based on Jeffries--Matusita (JM) distance
& \begin{itemize}
\item Improve classification accuracy
\item Classify hyperspectral images adequately
\item Ease of interpretation of urban classes
\end{itemize}
& \begin{itemize}
\item Confusion between road and bare soil classes
\item Instability and complexity in the structure and parameters of the binary tree SVM
\end{itemize}
& \begin{itemize}
\item \cite{du2012}
\end{itemize} \\
\midrule
Monitoring changes in impervious surfaces
& MRGU
& \begin{itemize}
\item Integrate multiple endmember spectral mixture analysis (MESMA), analysis of regression residuals, spatial statistics (Getis Ord) with Moran’s I, and urban growth theories in an effective manner
\item Quantify and identify the magnitude of impervious surface changes and their spatial distribution
\end{itemize}
& \begin{itemize}
\item Not universally applicable due to specific behaviour of maximum noise fraction (MNF)
\item Difficult to use for quantifying changes in urban centres
\item Performance (regression residuals) is subject to the structure of urban regions
\end{itemize}
& \begin{itemize}
\item \cite{shahtahmassebi2016}
\end{itemize} \\
\midrule
Change detection
& Unsupervised Neural network and feature transformation
& \begin{itemize}
\item Deep architecture (SDAE) for better representation of the relationships between feature- and pixel-pair
\item Mapping based FCA and function learning to highlight change in a robust manner
\item Denoising autoencoder
\end{itemize}
& \begin{itemize}
\item Uncertainty in the feature-pair
\item High computational cost
\item Complexity structures of stacked denoising autoencoders (SDAE)
\item Many constraints are required to extract useful features
\end{itemize}
& \begin{itemize}
\item \cite{zhang2016}
\end{itemize} \\
\midrule
Water quality, and sustainable water resources management
& Integrated data fusion and mining (IDFM) and artificial neural network (ANN)
& \begin{itemize}
\item A near real-time monitoring and the early warning system
\item Efficiency
\item Forecasting reliability
\item Potential for local adoption
\end{itemize}
& \begin{itemize}
\item Prediction accuracy may be effected by uncertainties in the fused data
\item A large number of variables is required to overcome the uncertainty
\item Not applicable for regional meteorology parameters
\end{itemize}
& \begin{itemize}
\item \cite{imen2015}
\end{itemize} \\
\midrule
Air temperature estimation
& SVM
& \begin{itemize}
\item Fully automated method
\item SVM regression is robust
\item Regression errors can be modelled at the pixel level, improving accuracy estimation
\end{itemize}
& \begin{itemize}
\item Requires expert users to apply SVM
\item Does not work well under non cloud-free conditions and require in situ measurements
\item Regression error distribution is insufficient
\end{itemize}
& \begin{itemize}
\item \cite{moser2015}
\end{itemize} \\
\midrule
Fine-scale population estimation for urban management, emergency response and epidemiological
& \begin{itemize}
\item RF
\item Linear regression modelling
\end{itemize}
& \begin{itemize}
\item Able to classify building types and extract their footprints in the heterogeneous urban areas
\item Improved classification accuracy
\item Ease of adoption
\end{itemize}
& \begin{itemize}
\item Subject to the accuracy of the selected morphology filter
\item Use of large numbers of metrics and variables for building type classification
\item Building background metrics do not show its advantage in the block classification
\item Classification uncertainty for non-residential buildings
\end{itemize}
& \begin{itemize}
\item \cite{xie2015}
\end{itemize} \\
\midrule
Renewable energy and urban feature extraction
& Shadow detection and building geometry identification
& \begin{itemize}
\item Easy to apply
\item Sufficient to generate 3D model of urban buildings
\item Reliable analysis of the solar energy potential
\item Identify the availability of 3D surfaces
\item Flexibility and feasibility
\end{itemize}
& \begin{itemize}
\item Not fully automated
\item Not suitable for dense urban areas
\item Sensitive to the quality of satellite images
\end{itemize}
& \begin{itemize}
\item \cite{kadhim2015a}
\end{itemize} \\
\midrule
Impervious surfaces estimation
& \begin{itemize}
\item SVM
\item RF
\end{itemize}
& \begin{itemize}
\item Increased classification accuracy
\item Does not depend on combinations of features
\item Data can be fused to optimise parameters efficiently
\item Ease of application
\end{itemize}
& \begin{itemize}
\item Needs many texture matrices
\item Inability to handle the confusion in shaded areas and bare soil
\item Over-fitting
\end{itemize}
& \begin{itemize}
\item \cite{zhang2015}
\end{itemize} \\
\end{longtable}
\end{document}