MathematicalEngineeringDeepLearning/tikz/figure2-9-error_modelcomplexity.tex at master · yoninazarathy/MathematicalEngineeringDeepLearning · GitHub

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
\documentclass[tikz, border=50pt]{standalone}

\usepackage{tikz}
\usepackage{geometry, amsmath}
\usepackage{medl_colors}
\usepackage{pgfplots}

\usetikzlibrary{shapes.multipart, shapes.geometric, arrows.meta}
\usetikzlibrary{matrix,decorations.pathreplacing, calc, positioning,fit}
\usetikzlibrary{positioning}

\pgfplotsset{compat=1.17}

\begin{document}
\begin{tikzpicture}

\newcommand\medlsquare[7]
{
    \addplot[thick, draw=gray!20,fill=#6] coordinates { (#1, #2) (#1+#3, #2) (#1+#3, #2+#4) (#1, #2+#4) (#1, #2)} node[] {};
    \path[draw, Triangle-Triangle] ([yshift=-0.4cm]#1+#3, #2+#4) -- ([yshift=-0.4cm]#1, #2+#4) node[midway,anchor=center, fill=#6] {#7};
}

\begin{axis}[
    axis line style = ultra thick,
    xlabel={\Large Model Complexity},
    ylabel={\Large Error},
    xticklabels={},
    yticklabels={},
    ytick style={draw=none},
    xtick style={draw=none},
    x label style={at={(.6,-.03)}},
    xmin = 0, xmax = 22,
    ymin = 0.1, ymax = 14,
    x=22,
    y=14,
    legend cell align=left,
    legend style={at={(.58,.95)},column sep=10pt},
    axis lines* = left
]

\medlsquare{1}{.2}{5}{14}{A}{fill-light-green}{\Large Underfitting}
\medlsquare{14}{.2}{7}{14}{B}{fill-light-red}{\Large Overfitting}

\addplot[thick, loosely dashed] coordinates {(7.7, .4) (7.7, 6)} node[align=center] at (8.1,7) {\Large Optimal Model};

\path (1.5,13) edge [out=280,in=200,draw=border-red, line width=.7mm] node [pos=.85, above=.8] {\LARGE $\tilde{E}_{\text{unseen}}$} (21,11);

\path (1.5,13) edge [out=280,in=175,draw=border-green, line width=.7mm] node [pos=.9, above=.5] {\LARGE $\tilde{E}_{\text{train}}$} (21,1);

\end{axis}
\end{tikzpicture}
\end{document}