[Tex/LaTex] Recurrent neural network with tikz

tikz-pgf

I would like to draw one curve arrow from the output to the input. Could anyone please help me? thank you beforehand

\documentclass{article}

\usepackage{tikz}
\begin{document}
\pagestyle{empty}

\def\layersep{2.5cm}

\begin{tikzpicture}[shorten >=1pt,->,draw=black!50, node distance=\layersep]
    \tikzstyle{every pin edge}=[<-,shorten <=1pt]
    \tikzstyle{neuron}=[circle,fill=black!25,minimum size=17pt,inner sep=0pt]
    \tikzstyle{input neuron}=[neuron, fill=green!50];
    \tikzstyle{output neuron}=[neuron, fill=red!50];
    \tikzstyle{hidden neuron}=[neuron, fill=blue!50];
    \tikzstyle{annot} = [text width=4em, text centered]

    % Draw the input layer nodes
    \foreach \name / \y in {1,...,4}
    % This is the same as writing \foreach \name / \y in {1/1,2/2,3/3,4/4}
        \node[input neuron, pin=left:Input \#\y] (I-\name) at (0,-\y) {};

    % Draw the hidden layer nodes
    \foreach \name / \y in {1,...,5}
        \path[yshift=0.5cm]
            node[hidden neuron] (H-\name) at (\layersep,-\y cm) {};

    % Draw the output layer node
    \node[output neuron,pin={[pin edge={->}]right:Output}, right of=H-3] (O) {};

    % Connect every node in the input layer with every node in the
    % hidden layer.
    \foreach \source in {1,...,4}
        \foreach \dest in {1,...,5}
            \path (I-\source) edge (H-\dest);

    % Connect every node in the hidden layer with the output layer
    \foreach \source in {1,...,5}
        \path (H-\source) edge (O);

    % Annotate the layers
    \node[annot,above of=H-1, node distance=1cm] (hl) {Hidden layer};
    \node[annot,left of=hl] {Input layer};
    \node[annot,right of=hl] {Output layer};
\end{tikzpicture}
% End of code
\end{document}

Best Answer

I guess, that you like to obtain something similar:

but I'm not sure (you not defined where is your "input")

\documentclass{article}
\usepackage{tikz}

\begin{document}
\pagestyle{empty}
\usetikzlibrary{calc}

\def\layersep{2.5cm}

    \begin{tikzpicture}[shorten >=1pt,->,draw=black!50, 
    node distance = \layersep,
every pin edge/.style = {<-,shorten <=1pt},
        neuron/.style = {circle,fill=black!25,minimum size=17pt,inner sep=0pt},
  input neuron/.style = {neuron, fill=green!50},
 output neuron/.style = {neuron, fill=red!50},
 hidden neuron/.style = {neuron, fill=blue!50},
         annot/.style = {text width=4em, text centered},
                         ]
% Draw the input layer nodes
\foreach \name / \y in {1,...,4}
% This is the same as writing \foreach \name / \y in {1/1,2/2,3/3,4/4}
    \node[input neuron, pin=left:Input \#\y] (I-\name) at (0,-\y) {};

% Draw the hidden layer nodes
\foreach \name / \y in {1,...,5}
    \path[yshift=0.5cm]
        node[hidden neuron] (H-\name) at (\layersep,-\y cm) {};

% Draw the output layer node
\node[output neuron,pin={[pin edge={->}]right:Output}, right of=H-3] (O) {};

% Connect every node in the input layer with every node in the hidden layer.
\foreach \source in {1,...,4}
    \foreach \dest in {1,...,5}
        \path (I-\source) edge (H-\dest);

% Connect every node in the hidden layer with the output layer
\foreach \source in {1,...,5}
    \path (H-\source) edge (O);

% Annotate the layers
\node[annot,above of=H-1, node distance=1cm] (hl) {Hidden layer};
\node[annot,left of=hl] {Input layer};
\node[annot,right of=hl] {Output layer};

% curve from output to input
\draw[line width=1mm, gray,->]  
    ([xshift=21mm] O.east) .. controls + (2,-1) and + (4,0) .. ([yshift=-7mm] H-5.south)
                           .. controls + (-5,0) and + (-2,-1) .. ([xshift=-27mm]$(I-1)!0.5!(I-4)$);
    \end{tikzpicture}
% End of code
\end{document}

Addendum: From OP comment follows that my guessing what curve should connect, was wrong (sorry, but my crystal ball is on strike :( ). Even now I'm not sure, if the curve lie as OP desire. However, the principle is in the first and this case the same and if also now the curve is not exactly where should be, only is necessary to accordingly change the start and end coordinates of the curve.

In version below are in comparison to the first one changed arrow type. For it is now used arrows.meta and bending library which enable more smooth positioned arrowhead with option flex. Also the curve, since now cross texts at inputs, is drawn on background layer. For this serves libraries backgrounds and scopes:

\documentclass{article}
\usepackage{tikz}

\begin{document}
\pagestyle{empty}
\usetikzlibrary{arrows.meta, bending, backgrounds, scopes}

\def\layersep{2.5cm}

    \begin{tikzpicture}[shorten >=1pt,->,draw=black!50, 
    node distance = \layersep,
every pin edge/.style = {<-,shorten <=1pt},
        neuron/.style = {circle,fill=black!25,minimum size=17pt,inner sep=0pt},
  input neuron/.style = {neuron, fill=green!50},
 output neuron/.style = {neuron, fill=red!50},
 hidden neuron/.style = {neuron, fill=blue!50},
         annot/.style = {text width=4em, text centered},
                         ]
% Draw the input layer nodes
\foreach \name / \y in {1,...,4}
% This is the same as writing \foreach \name / \y in {1/1,2/2,3/3,4/4}
    \node[input neuron, pin=left:Input \#\y] (I-\name) at (0,-\y) {};

% Draw the hidden layer nodes
\foreach \name / \y in {1,...,5}
    \path[yshift=0.5cm]
        node[hidden neuron] (H-\name) at (\layersep,-\y cm) {};

% Draw the output layer node
\node[output neuron,pin={[pin edge={->}]right:Output}, right of=H-3] (O) {};

% Connect every node in the input layer with every node in the hidden layer.
\foreach \source in {1,...,4}
    \foreach \dest in {1,...,5}
        \path (I-\source) edge (H-\dest);

% Connect every node in the hidden layer with the output layer
\foreach \source in {1,...,5}
    \path (H-\source) edge (O);

% Annotate the layers
\node[annot,above of=H-1, node distance=1cm] (hl) {Hidden layer};
\node[annot,left of=hl] {Input layer};
\node[annot,right of=hl] {Output layer};

% curve from output to input
\scoped[on background layer]
\draw[line width=1mm, gray!50, shorten >=1mm, shorten <=1mm, -{Latex[flex]}]  
    (O.east) .. controls + (1,-0.5) and + (3.5,0) .. ([yshift=-7mm] H-5.south)
             .. controls + (-6,0) and + (-2,-0.5) .. (I-1);
    \end{tikzpicture}
% End of code
\end{document}

Related Solutions

[Tex/LaTex] How to define the default vertical distance between nodes

I don't really get the question so I hope this is what you wanted. If you include a full document (such that we copy paste and see the problem on our systems) things are much more easier.

Here, you can change the default setting within a scope but your block style had a node distance which was resetting every time it is issued. I've made it 2mm such that we can see the difference easier.

\documentclass[tikz]{standalone}
\usetikzlibrary{arrows,shapes.geometric,positioning}

\begin{document}
\begin{tikzpicture}[decision/.style={diamond, draw, text width=4.5em, text badly centered, node distance=3.5cm, inner sep=0pt},
                    block/.style   ={rectangle, draw, text width=6em, text centered, rounded corners, minimum height=4em, minimum height=2em},
                    cloud/.style   ={draw, ellipse, minimum height=2em},
                    line/.style    ={draw,-latex'},
                    node distance = 1cm, 
                    auto]
    \node [block] (1st) {1st};
    \node [block, right= of 1st]  (2nd1) {2nd1};
\begin{scope}[node distance=2mm and 10mm]%Here we change it for everything inside this scope
    \node [block, above= of 2nd1] (2nd2) {2nd2};
    \node [block, below= of 2nd1] (2nd3) {2nd3};
    \node [block, right= of 2nd1] (3rd1) {3rd1};
    \node [block, above= of 3rd1] (3rd2) {3rd2};
    \node [block, above= of 3rd2] (3rd3) {3rd3};
\end{scope}
    \node [block, below= of 3rd1] (3rd4) {3rd4};
    \node [block, below= of 3rd4] (3rd5) {3rd5};

    \path [line] (1st) -- (2nd1);
    \path [line] (2nd1) -- (2nd2);
    \path [line] (2nd1) -- (2nd3);
    \path [line] (2nd2) -- (3rd3);
    \path [line] (2nd1) -- (3rd1);
    \path [line] (1st) -- (2nd1);
\end{tikzpicture}
\end{document}

enter image description here

[Tex/LaTex] TikZ: Drawing an arc from an intersection to an intersection

A solution which allows to draw intersection segments of any two intersections is available as tikz library fillbetween.

This library works as general purpose tikz library, but it is shipped with pgfplots and you need to load pgfplots in order to make it work:

\documentclass{standalone}

\usepackage{tikz}
\usepackage{pgfplots}
\usetikzlibrary{fillbetween}

\begin{document}

\begin{tikzpicture}

\draw [name path=red,red] (120:1.06) circle (1.9);
%\draw [name path=yellow,yellow] (0:1.06) circle (2.12);
\draw [name path=green,green!50!black] (0:0.77) circle (2.41);
\draw [name path=blue,blue] (0:0) circle (1.06);

% substitute this temp path by `\path` to make it invisible:
\draw[name path=temp1, intersection segments={of=red and blue,sequence=L1}];
\draw[red,-stealth,ultra thick, intersection segments={of=temp1 and green,sequence=L3}];

\end{tikzpicture}

\end{document}

enter image description here

The key intersection segments is described in all detail in the pgfplots reference manual section "5.6.6 Intersection Segment Recombination"; the key idea in this case is to

create a temporary path temp1 which is the first intersection segment of red and blue, more precisely, it is the first intersection segment in the Left argument in red and blue : red. This path is drawn as thin black path. Substitute its \draw statement by \path to make it invisible.
Compute the desired intersection segment by intersecting temp1 and green and use the correct intersection segment. By trial and error I figured that it is the third segment of path temp1 which is written as L3 (L = left argument in temp1 and green and 3 means third segment of that path).

The argument involves some trial and error because fillbetween is unaware of the fact that end and startpoint are connected -- and we as end users do not see start and end point.

Note that you can connect these path segments with other paths. If such an intersection segment should be the continuation of another path, use -- as before the first argument in sequence. This allows to fill paths segments:

\documentclass{standalone}

\usepackage{tikz}
\usepackage{pgfplots}
\usetikzlibrary{fillbetween}

\begin{document}

\begin{tikzpicture}

\draw [name path=red,red] (120:1.06) circle (1.9);
%\draw [name path=yellow,yellow] (0:1.06) circle (2.12);
\draw [name path=green,green!50!black] (0:0.77) circle (2.41);
\draw [name path=blue,blue] (0:0) circle (1.06);

% substitute this temp path by `\path` to make it invisible:
\draw[name path=temp1, intersection segments={of=red and blue,sequence=L1}];
\draw[red,fill=blue,-stealth,ultra thick, intersection segments={of=temp1 and green,sequence=L3}]  
    [intersection segments={of=temp1 and green, sequence={--R2}}]
;

\end{tikzpicture}

\end{document}

enter image description here

Best Answer

Related Solutions

[Tex/LaTex] How to define the default vertical distance between nodes

[Tex/LaTex] TikZ: Drawing an arc from an intersection to an intersection

Related Question