我正在尝试生成一组包含 3 个数字的条形图并成功完成，输出如下所示。
问题：我希望前两个条形图（a 和 b）处于相同状态，但对于第三个条形图（（c）Q 损失和 P 损失），我如何在 X 轴上为每个标签添加两个条形图，并为该图表添加单独的图例？重申一下，我想要 3-6 的两个条形图、6-12 的两个条形图和 12-24 的两个条形图，并为第三个图表添加单独的图例。另外，有人能帮我减少条形图（a、b、c）之间的空间吗？
输出：
乳胶代码：
\documentclass[10pt,journal,compsoc,xcolor={dvipsnames}]{IEEEtran}
%[tikz]{standalone}
% If IEEEtran.cls has not been installed into the LaTeX system files,
% manually specify the path to it like:
% \documentclass[10pt,journal,compsoc]{../sty/IEEEtran}
%\usepackage{showframe}
%\renewcommand\ShowFrameLinethickness{0.15pt}
%\renewcommand*\ShowFrameColor{\color{red}}
%---------------------------------------------------------------%
\usepackage{pgfplots}
\usepgfplotslibrary{groupplots}
%\usepackage[dvipsnames]{xcolor}
\pgfplotsset{compat=1.18}
\usepackage{pgfplotstable}
\usepackage{graphicx}
\usepackage{subfig}
\usepackage{multirow}
\usepackage{lipsum}
\usepackage{xspace}
\usetikzlibrary{patterns}
\usepackage{amsmath}
\usepackage{amsfonts}
\usepackage{amssymb}
\newcommand{\R}{\mathbb{R}}
\newcommand*\circled[1]{\tikz[baseline=(char.base)]{
\node[shape=circle,draw,inner sep=0.7pt] (char) {#1};}}
\begin{document}
\begin{figure}[!htpb]
\centering
\begin{tikzpicture}
\begin{groupplot}[
legend columns=-1,
%bar width=0.22, % Width of the bar
%, % Distance between the centers of the bars
symbolic x coords={3-6, 6-12, 12-24},
enlarge x limits={abs=0.4cm},
%width=0.25\textwidth,
xtick={3-6, 6-12, 12-24},
%height=0.6\textheight,
ymin=0,
ymax=7,
xticklabel style={font=\tiny},
nodes near coords,
every node near coord/.append style={font=\tiny},
nodes near coords align={vertical},
xlabel={\tiny Agents~$(N \to \ 2N)$},
ylabel={\tiny Growth rate~($N\times$)},
% The distance between the center of the first bar and the left edge
%enlarge y limits=false
legend entries={{\color{olive}{\tiny Random}},{\color{orange}{\tiny ++Cost}},{\color{teal}{\tiny ++FTE}}},
legend to name=CombinedLegendBar,
footnotesize,
ybar legend,
width=0.4\columnwidth,
legend image code/.code={%
\draw[#1] (0cm,-0.1cm) rectangle (0.6cm,0.1cm);
},
% area legend, % This is the alternate option
group style={
group size=3 by 1,
xlabels at=edge bottom,
ylabels at=edge left}]
\nextgroupplot[title={\tiny (a) Mini-batch sampling}]
\addplot[ybar, pattern color=black, pattern=crosshatch dots] coordinates { (3-6, 3.5)};
\addplot[ybar, pattern color=gray, pattern=checkerboard] coordinates { (6-12, 4.1)};
\addplot[ybar, pattern color=gray, pattern=north east lines] coordinates { (12-24, 4.5)};
\nextgroupplot[title={\tiny (b) Target Q Calculation}]
\addplot[ybar, pattern color=black, pattern=crosshatch dots] coordinates { (3-6, 4.1)};
\addplot[ybar, pattern color=gray, pattern=checkerboard] coordinates { (6-12, 4.6)};
\addplot[ybar,pattern color=gray, pattern=north east lines] coordinates { (12-24, 4.6)};
\nextgroupplot[title={\tiny (c) Q loss \& P loss}]
` \addplot[ybar, pattern color=black, pattern=crosshatch dots] coordinates { (3-6, 3.05)};
\addplot[ybar, pattern color=gray, pattern=checkerboard] coordinates { (6-12,3.23)};
\addplot[ybar, pattern color=gray, pattern=north east lines] coordinates { (12-24, 3.42)};
\end{groupplot}
\end{tikzpicture}
\vspace{-\baselineskip}
\caption{Update all trainers sub-functions super-linear growth rate for all the three workloads when the number of agents are scaled from $N-2N$.}% The environment is Cooperative Navigation (simple spread).}
\vspace{-\baselineskip}
\label{Figure4}
\end{figure}
\end{document}
有人能帮我解决这个问题吗？
谢谢！
答案1
- 显然，我第一次尝试回答问题时误解了它......
- 从@Torbjørn T. 的精彩回答中学习，现在我知道，主要问题是在最后一张图 P 中添加一组蓝色的条形图。
- 这相对容易做到：
- Q 组（黑色）的条形图向左小幅移动
- 添加新的条形图 P（蓝色），并将其向右移动相同的量
- 在新的 MWE 中保留了答案定义第一版的图表样式。
- 现在使用模式
patterns.meta库。
\documentclass[journal,compsoc,xcolor={dvipsnames}]{IEEEtran}
\usepackage{pgfplots}
\usepgfplotslibrary{groupplots}
\pgfplotsset{compat=1.18}
\usetikzlibrary{patterns, patterns.meta}
\begin{document}
\begin{figure}[ht]
\raggedright
\begin{tikzpicture}[
pattern-A/.style = {pattern={Dots[angle=45, distance=2pt,xshift=1pt]}},
pattern-B/.style = {pattern={Hatch[angle=45,distance=2pt,xshift=1pt]}},
pattern-C/.style = {pattern={Lines[angle=45,distance=2pt]}},
%
pattern-AA/.style = {pattern-A, pattern color=blue},
pattern-BB/.style = {pattern-B, pattern color=blue},
pattern-CC/.style = {pattern-C, pattern color=blue},
]
\begin{groupplot}[
group style={
group size=3 by 1,
ylabels at=edge left,
horizontal sep=4mm,
},
width=0.44\linewidth,
enlarge x limits=0.4,
%
title style = {yshift=-1ex, font=\scriptsize, align=center},
xlabel={Agents~$(N \to \ 2N)$},
ylabel={Growth rate~($N\times$)},
label style = {font=\scriptsize},
%
ybar, % <---
/pgf/bar width=2.4mm, % <--- common set, overwritten in third \nextgroupplot
/pgf/bar shift=0pt, % <--- common set, overwritten in third \nextgroupplot
%
symbolic x coords={3-6, 6-12, 12-24},
ticklabel style={font=\scriptsize},
ymin=0, ymax=7,
%
nodes near coords,
every node near coord/.append style={font=\tiny},
nodes near coords align={vertical},
%
legend image code/.code={\draw[#1] (0cm,-1mm) rectangle (4mm,1mm);},
legend style={legend columns=-1,
font=\footnotesize,
/tikz/every even column/.append style={column sep=1em}
}
]
\nextgroupplot[title={(a) Mini-batch\\ sampling}]
\addplot[pattern-A] coordinates { (3-6, 3.5)};
\addplot[pattern-B] coordinates { (6-12, 4.1)};
\addplot[pattern-C] coordinates { (12-24, 4.5)};
\nextgroupplot[title={(b) Target Q\\ Calculation}]
\addplot[pattern-A] coordinates { (3-6, 4.1)};
\addplot[pattern-B] coordinates { (6-12, 4.6)};
\addplot[pattern-C] coordinates { (12-24, 4.6)};
\nextgroupplot[title={(c) Q loss\\ \& \textcolor{blue}{P loss}},
nodes near coords style={rotate=90, anchor=west},
legend to name=CombinedLegendBar,
]
\addplot[pattern-A,bar shift=-1.4mm] coordinates { (3-6, 3.05)};
\addplot[pattern-B,bar shift=-1.4mm] coordinates { (6-12, 3.23)};
\addplot[pattern-C,bar shift=-1.4mm] coordinates { (12-24,3.42)};
%
\addplot[pattern-AA,bar shift=+1.4mm] coordinates { (3-6,4.05)};
\addplot[pattern-BB,bar shift=+1.4mm] coordinates { (6-12,4.23)};
\addplot[pattern-CC,bar shift=+1.4mm] coordinates { (12-24,4.42)};
\legend{Qa, Qb, Qc, Pa, Pb, Pc}
\end{groupplot}
\node [above=2\baselineskip] at (group c2r1.north) {\ref{CombinedLegendBar}};
\end{tikzpicture}
\caption{Update all trainers sub-functions super-linear growth rate for all the three workloads when the number of agents are scaled from $N-2N$.}% The environment is Cooperative Navigation (simple spread).}
\label{Figure4}
\end{figure}
\end{document}
答案2
可能需要进行一些额外的调整，我将图例放在了图上方，并添加了占位符条目。轴内实际上没有空间放置图例。
主要的变化是使用ybar作为选项的groupplot而不是每个的\addplot，并使用手动栏移动来正确放置栏。
horizontal sep在group style设置轴之间的空间。
\documentclass[10pt,journal,compsoc,xcolor={dvipsnames}]{IEEEtran}
\usepackage{pgfplotstable}
\usepgfplotslibrary{groupplots}
\pgfplotsset{compat=1.18}
\usetikzlibrary{patterns}
\begin{document}
\begin{figure}[!htpb]
\centering
\begin{tikzpicture}
\begin{groupplot}[
legend columns=-1,
%bar width=0.22, % Width of the bar
%, % Distance between the centers of the bars
symbolic x coords={3-6, 6-12, 12-24},
enlarge x limits=0.3,
%width=0.25\textwidth,
xtick={3-6, 6-12, 12-24},
%height=0.6\textheight,
ymin=0,
ymax=7,
xticklabel style={font=\tiny},
nodes near coords,
every node near coord/.append style={font=\tiny},
nodes near coords align={vertical},
xlabel={\tiny Agents~$(N \to \ 2N)$},
ylabel={\tiny Growth rate~($N\times$)},
% The distance between the center of the first bar and the left edge
%enlarge y limits=false
footnotesize,
ybar=-\pgfplotbarwidth,
width=0.45\columnwidth,
legend image code/.code={%
\draw[#1] (0cm,-0.1cm) rectangle (0.2cm,0.2cm);
},
legend style={cells={font=\scriptsize}},
% area legend, % This is the alternate option
group style={
group name=foo,
group size=3 by 1,
xlabels at=edge bottom,
ylabels at=edge left,
horizontal sep=4.5mm
},
title style={font=\tiny}]
\nextgroupplot[title={(a) Mini-batch sampling}]
\addplot[pattern color=black, pattern=crosshatch dots] coordinates { (3-6, 3.5)};
\addplot[pattern color=gray, pattern=checkerboard] coordinates { (6-12, 4.1)};
\addplot[pattern color=gray, pattern=north east lines] coordinates { (12-24, 4.5)};
\nextgroupplot[title={(b) Target Q Calculation}]
\addplot[pattern color=black, pattern=crosshatch dots] coordinates { (3-6, 4.1)};
\addplot[pattern color=gray, pattern=checkerboard] coordinates { (6-12, 4.6)};
\addplot[ybar,pattern color=gray, pattern=north east lines] coordinates { (12-24, 4.6)};
\nextgroupplot[
title={(c) Q loss \& P loss},
bar width=8pt,
legend to name=CombinedLegendBar,
]
\pgfmathsetlengthmacro{\MyBarshift}{0.5*\pgfplotbarwidth+1pt}
\addplot[bar shift=-\MyBarshift,pattern color=black, pattern=crosshatch dots] coordinates { (3-6, 3.05)};
\addplot[bar shift=-\MyBarshift,pattern color=gray, pattern=checkerboard] coordinates { (6-12,3.23)};
\addplot[bar shift=-\MyBarshift,pattern color=gray, pattern=north east lines] coordinates { (12-24, 3.42)};
% second set of bars
\addplot[bar shift=\MyBarshift,pattern color=blue, pattern=crosshatch dots] coordinates { (3-6, 4.05)};
\addplot[bar shift=\MyBarshift,pattern color=blue, pattern=checkerboard] coordinates { (6-12,4.23)};
\addplot[bar shift=\MyBarshift,pattern color=blue, pattern=north east lines] coordinates { (12-24, 4.42)};
\legend{Qa,Qb,Qc,Pa,Pb,Pc}
\end{groupplot}
\node [above=12pt] at (foo c2r1.north) {\ref{CombinedLegendBar}};
\end{tikzpicture}
\vspace{-\baselineskip}
\caption{Update all trainers sub-functions super-linear growth rate for all the three workloads when the number of agents are scaled from $N-2N$.}% The environment is Cooperative Navigation (simple spread).}
\vspace{-\baselineskip}
\label{Figure4}
\end{figure}
\end{document}
答案3
这些条形图是通过
pic依赖于三个参数（高度、标签和填充模式）的元素创建的。尺寸在开始时作为全局变量进行控制；您可以根据需要修改它们。
代码
\documentclass[11pt, margin=1cm]{standalone}
\usepackage{tikz}
\usetikzlibrary{math, patterns}
\begin{document}
\tikzmath{
real \w, \dw, \globalW, \globalH, \g1, \g2, \g3;
\w = 1.5; % width of a bar
\dw = .4; % half inbetween two bars
\globalW = 3*\w +6*\dw;
\globalH = 6;
\g1 = 0; % position of the first graph
\g2 = \globalW +1.3*\w;
\g3 = 2*\globalW +4*\w;
}
\tikzset{
pics/bar/.style n args={3}{% height, label, pattern
code={%
\draw[pattern=#3] (0, 0) rectangle (\w, #1/7*\globalH)
++ (-\w/2, 1ex) node[above, scale=.8] {#1};
\path (\w/2, -1ex) node[below, scale=.8] {#2};
}
}
}
\begin{tikzpicture}[every node/.style={scale=.9}]
\begin{scope}[xshift={\g1 cm}]
\path (\globalW/2, \globalH) node[above] {(a) Mini-batch sampling};
\draw (0, 0) -- ++(\globalW, 0) ++(-\globalW/2, -4ex)
node[below] {Agents $(N \to 2N)$};
\draw (0, 0) -- ++(0, \globalH) ++(-3em, -\globalH/2)
node[rotate=90] {Growth rate $(N\times)$};
\foreach \j in {0, 2, 4, 6}{%
\draw (0, \j/7*\globalH) -- ++(-1ex, 0) node[left] {$\j$};
}
\path
(0, 0) ++(\dw, 0) pic {bar={3.5}{3-6}{dots}}
++(\w +2*\dw, 0) pic {bar={4.1}{6-12}{crosshatch}}
++(\w +2*\dw, 0) pic {bar={4.5}{12-24}{north east lines}};
\end{scope}
\begin{scope}[xshift={\g2 cm}]
\path (\globalW/2, \globalH) node[above] {(b) Target Q Calculation};
\draw (0, 0) -- ++(\globalW, 0) ++(-\globalW/2, -4ex)
node[below] {Agents $(N \to 2N)$};
\draw (0, 0) -- ++(0, \globalH) ++(-3em, -\globalH/2);
\foreach \j in {0, 2, 4, 6}{%
\draw (0, \j/7*\globalH) -- ++(-1ex, 0) node[left] {$\j$};
}
\path
(0, 0) ++(\dw, 0) pic {bar={4.1}{3-6}{dots}}
++(\w +2*\dw, 0) pic {bar={4.6}{6-12}{crosshatch}}
++(\w +2*\dw, 0) pic {bar={4.6}{12-24}{north east lines}};
\end{scope}
\begin{scope}[xshift={\g3 cm}]
\path (\globalW/2, \globalH) node[above] {(c) Q loss \& P loss};
\draw (0, 0) -- ++(\globalW, 0) ++(-\globalW/2, -4ex)
node[below] {Agents $(N \to 2N)$};
\draw (0, 0) -- ++(0, \globalH) ++(-3em, -\globalH/2)
node[rotate=90] {Growth rate $(N\times)$};
\foreach \j in {0, 2, 4, 6}{%
\draw (0, \j/7*\globalH) -- ++(-1ex, 0) node[left] {$\j$};
}
\path
(0, 0) ++(\dw, 0) pic {bar={3.05}{3-6}{dots}}
++(\w +2*\dw, 0) pic {bar={3.25}{6-12}{crosshatch}}
++(\w +2*\dw, 0) pic {bar={3.42}{12-24}{north east lines}};
\end{scope}
\end{tikzpicture}
\end{document}