我对感兴趣适当地重现上图。红色标记不是图的一部分，只是为了识别块。我对以下方面特别感兴趣。
- 块 A 的上边框应与块 B1 的上边框高度相同；其下边框应与块 BK 的下边框高度相同。块 C 也是如此。
- 区块 A 与区块 Bi (i=1,...,K) 之间的分叉应该如上图所示，即有足够的空间容纳 $\vec p(\vec c)$ 并且呈这种角度形式。
- 在块内部，AI 希望在顶部、底部和中间放置文本。
- 理想情况下，我可以设置 B2 块和 BK 块之间的距离。
下面是我的尝试。我不喜欢的是：
- 要求 1 并未完全满足。我必须手动设置块 A 和 C 的高度，但框仍然未对齐（可能高达 的倍数
line width）。 - B2和BK之间的空白区域太多（两个看不见的小节点）。
- 块 A 内的间距是手工制作的。也许类似的东西
$p_1...$\\$p_2...$\vfill$\vdots$\vfill$p_K...$可以完成这项工作。
\documentclass[border=2mm]{standalone}
\usepackage[dvipsnames]{xcolor}
\usepackage{tikz,amsmath}
\usetikzlibrary{positioning}
\begin{document}
\begin{tikzpicture}[
node distance=0.2cm and 1.2cm,
small node/.style={rectangle, draw, minimum width=2.5cm, minimum height=1 cm, align=center, fill=lightgray},
large node/.style={small node, minimum height=5.8cm},
invisible node/.style={minimum size=0pt, inner sep=0pt, outer sep=0pt},
]
% Nodes.
\node[invisible node] (init) {};
\node[large node, right=of init, align=center] (a) {$p_1p(\vec c\vert\Omega_1)$\\$p_2p(\vec c\vert\Omega_2)$\\[1cm]$\vdots$\\[1cm]$p_Kp(\vec c\vert\Omega_K)$};
\node[invisible node, right=of a] (b) {};
\node[small node, right=of b , draw=none, fill=none] (c3) {};
\node[small node, above=of c3 ] (c2) {$\vec r_1^\top \vec p(\vec c)$};
\node[small node, above=of c2, ] (c1) {$\vec r_0^\top \vec p(\vec c)$};
\node[small node, below=of c3, draw=none, fill=none] (c4) {};
\node[small node, below=of c4 ] (c5) {$\vec r_K^\top \vec p(\vec c)$};
\node[large node, right=of c3] (d) {$\kappa^\ast = \arg\min_\kappa u_\kappa$};
\node[invisible node, right=of d] (e) {};
% Arrows.
\draw[->] (init) -- node[anchor=south, pos=0.5] {$\vec c$} (a);
\draw[ ] (a) -- node[anchor=south, pos=0.5] {$\vec p(\vec c)$} (b);
\draw[->] (b) |- node[anchor=south, pos=0.5] {} (c1);
\draw[->] (b) |- node[anchor=south, pos=0.5] {} (c2);
\draw[->] (b) |- node[anchor=south, pos=0.5] {} (c5);
\draw[->] (c1) -- node[anchor=south]{$u_0(\vec c)$} (c1 -| d.west);
\draw[->] (c2) -- node[anchor=south]{$u_1(\vec c)$} (c2 -| d.west);
\draw[->] (c5) -- node[anchor=south]{$u_K(\vec c)$} (c5 -| d.west);
\draw[->] (d) -- node[anchor=south] {$\Omega_{\kappa^\ast}$} (e);
\end{tikzpicture}
\end{document}
答案1
以下是一些想法：
首先绘制中间的框并用作大框的参考（通过
ext.positioning-plus和它的部分|和使得新框的总高度等于其他框的高度）。ofleftright左侧框的内容作为标签或
它们
dumb labels与相应的小盒子放置在相同的高度。horizontal vertical horizontal这些线由（即-|-）和绘制only horizontal first。设置各种
minimum width要求minimum height，以使节点看起来不那么紧凑，但左侧框也足够宽以容纳其标签。（可以使用自动方法，但这种方法更快。）
代码
\documentclass[tikz]{standalone}
\usetikzlibrary{
arrows.meta,
ext.paths.ortho,
ext.positioning-plus,
quotes,
}
\tikzset{
math nodes/.style={execute at begin node=$, execute at end node=$},
dumb label/.style args={[#1]#2}{label={[{tikz@label@post/.code=,anchor=center,#1}]#2}},
}
\begin{document}
\begin{tikzpicture}[
every box/.style={draw, fill=lightgray, minimum width=+6em},
small box/.style={every box, minimum height=+2em},
big box/.style={every box},
every label/.append style={align=center},
node distance=2mm and 4em,
every edge quotes/.append style=math nodes,
]
\node[small box] (c0) {$\vec r_0^\top \vec p(\vec c)$};
\node[small box] (c1) [below= of c0] {$\vec r_1^\top \vec p(\vec c)$};
\node[small box] (cK) [below=4em of c1] {$\vec r_K^\top \vec p(\vec c)$};
% ^^^ vertical distance between c1 and ck
\node[big box] (box1) [left=of |(c0)(cK),
label={[below]above:$p_1p(\vec c \,\vert \Omega_1)$ \\
$p_2p(\vec c\, \vert \Omega_2)$},
label={[above]below:$p_Kp(\vec c \,\vert \Omega_K)$},
]{$\vdots$};
\node[big box] (box2) [right=of |(c0)(cK)]{$\kappa^\ast = \arg\min_\kappa u_\kappa$};
\path[->] (box1.west) edge[<-, "\vec c"'] +(left:x_node_dist)
[horizontal vertical horizontal, ortho/ratio=.6666]
(box1) edge (c0)
edge["\vec p(\vec c)" very near start] (c1)
edge (cK)
[only horizontal first]
(c0) edge["u_0(\vec c)"] (box2)
(c1) edge["u_1(\vec c)"] (box2)
(cK) edge["u_\kappa(\vec c)"] (box2)
[line to] % back to --
(box2.east) edge["\Omega_{\kappa^\ast}"] +(right:x_node_dist)
;
\end{tikzpicture}
\begin{tikzpicture}[
every box/.style={draw, fill=lightgray, minimum width=+6em},
small box/.style={every box, minimum height=+2em},
big box/.style={every box},
every label/.append style={align=center},
node distance=2mm and 4em,
every edge quotes/.append style=math nodes,
]
\node[small box] (c0) {$\vec r_0^\top \vec p(\vec c)$};
\node[small box] (c1) [below= of c0] {$\vec r_1^\top \vec p(\vec c)$};
\node[small box] (cK) [below=4em of c1] {$\vec r_K^\top \vec p(\vec c)$};
% ^^^ vertical distance between c1 and ck
\node[big box] (box1) [left=of |(c0)(cK),
@/.style={dumb label={[at=(\tikzlastnode|-c#1), name=\tikzlastnode#1]%
$p_{#1}p(\vec c \,\vert \Omega_{#1})$}},
@/.list={0, 1, K},
]{};
\path (box11) -- node {$\vdots$} (box1K); % or at ($(box11)!.5!(box1K)$)
\node[big box] (box2) [right=of |(c0)(cK)]{$\kappa^\ast = \arg\min_\kappa u_\kappa$};
\path[->] (box1.west) edge[<-, "\vec c"'] +(left:x_node_dist)
[horizontal vertical horizontal, ortho/ratio=.6666]
(box1) edge (c0)
edge["\vec p(\vec c)" very near start] (c1)
edge (cK)
[only horizontal first]
(c0) edge["u_0(\vec c)"] (box2)
(c1) edge["u_1(\vec c)"] (box2)
(cK) edge["u_\kappa(\vec c)"] (box2)
[line to] % back to --
(box2.east) edge["\Omega_{\kappa^\ast}"] +(right:x_node_dist)
;
\end{tikzpicture}
\end{document}
输出
