Quellcodebibliothek Statistik Leitseite products/Sources/formale Sprachen/GAP/pkg/thelma/doc/   (Algebra von RWTH Aachen Version 4.15.1©)  Datei vom 3.2.2022 mit Größe 11 kB image not shown  

Quelle  chap4_mj.html   Sprache: HTML

 
 products/Sources/formale Sprachen/GAP/pkg/thelma/doc/chap4_mj.html


<?xml version="1.0" encoding="UTF-8"?>

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN"
         "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">

<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en">
<head>
<script type="text/javascript"
  src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.0/MathJax.js?config=TeX-AMS-MML_HTMLorMML">
</script>
<title>GAP (Thelma) - Chapter 4: Networks of Threshold Elements</title>
<meta http-equiv="content-type" content="text/html; charset=UTF-8" />
<meta name="generator" content="GAPDoc2HTML" />
<link rel="stylesheet" type="text/css" href="manual.css" />
<script src="manual.js" type="text/javascript"></script>
<script type="text/javascript">overwriteStyle();</script>
</head>
<body class="chap4"  onload="jscontent()">


<div class="chlinktop"><span class="chlink1">Goto Chapter: </span><a href="chap0_mj.html">Top</a>  <a href="chap1_mj.html">1</a>  <a href="chap2_mj.html">2</a>  <a href="chap3_mj.html">3</a>  <a href="chap4_mj.html">4</a>  <a href="chap5_mj.html">5</a>  <a href="chapBib_mj.html">Bib</a>  <a href="chapInd_mj.html">Ind</a>  </div>

<div class="chlinkprevnexttop"> <a href="chap0_mj.html">[Top of Book]</a>   <a href="chap0_mj.html#contents">[Contents]</a>    <a href="chap3_mj.html">[Previous Chapter]</a>    <a href="chap5_mj.html">[Next Chapter]</a>   </div>

<p id="mathjaxlink" class="pcenter"><a href="chap4.html">[MathJax off]</a></p>
<p><a id="X7BA69E617DE465FE" name="X7BA69E617DE465FE"></a></p>
<div class="ChapSects"><a href="chap4_mj.html#X7BA69E617DE465FE">4 <span class="Heading">Networks of Threshold Elements</span></a>
<div class="ContSect"><span class="tocline"><span class="nocss"> </span><a href="chap4_mj.html#X82EB5BE77F9F686A">4.1 <span class="Heading">Basic Operations</span></a>
</span>
<div class="ContSSBlock">
<span class="ContSS"><br /><span class="nocss">  </span><a href="chap4_mj.html#X780C027786A8E34B">4.1-1 NeuralNetwork</a></span>
<span class="ContSS"><br /><span class="nocss">  </span><a href="chap4_mj.html#X79B24B8F79161064">4.1-2 IsNeuralNetwork</a></span>
<span class="ContSS"><br /><span class="nocss">  </span><a href="chap4_mj.html#X7C93C473784D1676">4.1-3 OutputOfNeuralNetwork</a></span>
</div></div>
<div class="ContSect"><span class="tocline"><span class="nocss"> </span><a href="chap4_mj.html#X7BA69E617DE465FE">4.2 <span class="Heading">Networks of Threshold Elements</span></a>
</span>
<div class="ContSSBlock">
<span class="ContSS"><br /><span class="nocss">  </span><a href="chap4_mj.html#X786FB7237D875561">4.2-1 BooleanFunctionByNeuralNetwork</a></span>
<span class="ContSS"><br /><span class="nocss">  </span><a href="chap4_mj.html#X7C6E4221828E6609">4.2-2 BooleanFunctionByNeuralNetworkDASG</a></span>
</div></div>
</div>

<h3>4 <span class="Heading">Networks of Threshold Elements</span></h3>

<p>Not all Boolean functions can be realized by a single threshold element. However, all of them can be realized by a multi-layered network of threshold elements, with a number of threshold elements on a first layer and conjunction or a disjunction on the second layer. In this chapter we will decribe some functions regarding such networks.</p>

<p><a id="X82EB5BE77F9F686A" name="X82EB5BE77F9F686A"></a></p>

<h4>4.1 <span class="Heading">Basic Operations</span></h4>

<p>In this section we describe some operations, similar to the ones described in Section <a href="chap3_mj.html#X82EB5BE77F9F686A"><span class="RefLink">3.1</span></a>.</p>

<p><a id="X780C027786A8E34B" name="X780C027786A8E34B"></a></p>

<h5>4.1-1 NeuralNetwork</h5>

<div class="func"><table class="func" width="100%"><tr><td class="tdleft"><code class="func">‣ NeuralNetwork</code>( <var class="Arg">InnerLayer</var>, <var class="Arg">OuterLayer</var> )</td><td class="tdright">( function )</td></tr></table></div>
<p>For the list of threshold elements <code class="code">InnerLayer</code> and the Boolean variable <code class="code">OuterLayer</code>, which can be either <code class="code">true</code> (for disjunction), <code class="code">false</code> (for conjunction), or <code class="code">fail</code> (if there is only one layer) the function <code class="code">NeuralNetwork</code> returns a neural network built from this inputs.</p>


<div class="example"><pre>

<span class="GAPprompt">gap></span> <span class="GAPinput">te1:=ThresholdElement([1,1],1);</span>
< threshold element with weight vector [ 1, 1 ] and threshold 1 >
<span class="GAPprompt">gap></span> <span class="GAPinput">te2:=ThresholdElement([-1,-2],-2);</span>
< threshold element with weight vector [ -1, -2 ] and threshold -2 >
<span class="GAPprompt">gap></span> <span class="GAPinput">inner:=[te1,te2];</span>
[ < threshold element with weight vector [ 1, 1 ] and threshold 1 >,
  < threshold element with weight vector [ -1, -2 ] and threshold -2 > ]
<span class="GAPprompt">gap></span> <span class="GAPinput">nn:=NeuralNetwork(inner,false);</span>
< neural network with
2 threshold elements on inner layer and conjunction on outer level >
<span class="GAPprompt">gap></span> <span class="GAPinput">Display(last);</span>
Inner Layer:
[ [[ 1, 1 ], 1], [[ -1, -2 ], -2] ]
Outer Layer: conjunction
Neural Network realizes the function f :
Boolean function of 2 variables.
[ 0, 0 ] || 0
[ 0, 1 ] || 1
[ 1, 0 ] || 1
[ 1, 1 ] || 0
Sum of Products:[ 1, 2 ]

</pre></div>

<p><a id="X79B24B8F79161064" name="X79B24B8F79161064"></a></p>

<h5>4.1-2 IsNeuralNetwork</h5>

<div class="func"><table class="func" width="100%"><tr><td class="tdleft"><code class="func">‣ IsNeuralNetwork</code>( <var class="Arg">Obj</var> )</td><td class="tdright">( function )</td></tr></table></div>
<p>For the object <code class="code">Obj</code> the function <code class="code">IsNeuralNetwork</code> returns <code class="code">true</code> if <code class="code">Obj</code> is a neural network (see <code class="func">NeuralNetwork</code> (<a href="chap4_mj.html#X780C027786A8E34B"><span class="RefLink">4.1-1</span></a>)), and <code class="code">false</code> otherwise.</p>


<div class="example"><pre>

<span class="GAPprompt">gap></span> <span class="GAPinput">## Consider the neural network <C>nn</C> from the previous example.</span>
<span class="GAPprompt">gap></span> <span class="GAPinput">IsNeuralNetwork(nn);</span>
true

</pre></div>

<p><a id="X7C93C473784D1676" name="X7C93C473784D1676"></a></p>

<h5>4.1-3 OutputOfNeuralNetwork</h5>

<div class="func"><table class="func" width="100%"><tr><td class="tdleft"><code class="func">‣ OutputOfNeuralNetwork</code>( <var class="Arg">NNetwork</var> )</td><td class="tdright">( function )</td></tr></table></div>
<p>For the neural network <code class="code">NNetwork</code> the function <code class="code">OutputOfNeuralNetwork</code> returns the Boolean function, realized by <code class="code">NNetwork</code>.</p>


<div class="example"><pre>

<span class="GAPprompt">gap></span> <span class="GAPinput">f:=OutputOfNeuralNetwork(nn);</span>
< Boolean function of 2 variables >
<span class="GAPprompt">gap></span> <span class="GAPinput">Display(last);</span>
Boolean function of 2 variables.
[ 0, 0 ] || 0
[ 0, 1 ] || 1
[ 1, 0 ] || 1
[ 1, 1 ] || 0

</pre></div>

<p><a id="X7BA69E617DE465FE" name="X7BA69E617DE465FE"></a></p>

<h4>4.2 <span class="Heading">Networks of Threshold Elements</span></h4>

<p>In this section we consider the networks of threshold elements.</p>

<p><a id="X786FB7237D875561" name="X786FB7237D875561"></a></p>

<h5>4.2-1 BooleanFunctionByNeuralNetwork</h5>

<div class="func"><table class="func" width="100%"><tr><td class="tdleft"><code class="func">‣ BooleanFunctionByNeuralNetwork</code>( <var class="Arg">Func</var> )</td><td class="tdright">( function )</td></tr></table></div>
<p>For the Boolean function <code class="code">Func</code> the function <code class="code">BooleanFunctionByNeuralNetwork</code> returns a two-layered neural network, which realizes <code class="code">Func</code> (see <code class="func">NeuralNetwork</code> (<a href="chap4_mj.html#X780C027786A8E34B"><span class="RefLink">4.1-1</span></a>)). The realization of this function is based on the algorithm proposed in <a href="chapBib_mj.html#biBGecheRobotyshyn83">[GPR83]</a>.</p>


<div class="example"><pre>

<span class="GAPprompt">gap></span> <span class="GAPinput">x:=Indeterminate(GF(2),"x");;</span>
<span class="GAPprompt">gap></span> <span class="GAPinput">y:=Indeterminate(GF(2),"y");;</span>
<span class="GAPprompt">gap></span> <span class="GAPinput">z:=Indeterminate(GF(2),"z");;</span>
<span class="GAPprompt">gap></span> <span class="GAPinput">f:=PolynomialToBooleanFunction(x*y+z,3);</span>
< Boolean function of 3 variables >
<span class="GAPprompt">gap></span> <span class="GAPinput">nn:=BooleanFunctionByNeuralNetwork(f);</span>
< neural network with
2 threshold elements on inner layer and disjunction on outer level >
<span class="GAPprompt">gap></span> <span class="GAPinput">Display(last);</span>
Inner Layer:
[ [[ -1, -2, 4 ], 2], [[ 1, 2, -3 ], 3] ]
Outer Layer: disjunction
Neural Network realizes the function f :
Boolean function of 3 variables.
[ 0, 0, 0 ] || 0
[ 0, 0, 1 ] || 1
[ 0, 1, 0 ] || 0
[ 0, 1, 1 ] || 1
[ 1, 0, 0 ] || 0
[ 1, 0, 1 ] || 1
[ 1, 1, 0 ] || 1
[ 1, 1, 1 ] || 0
Sum of Products:[ 1, 3, 5, 6 ]

</pre></div>

<p><a id="X7C6E4221828E6609" name="X7C6E4221828E6609"></a></p>

<h5>4.2-2 BooleanFunctionByNeuralNetworkDASG</h5>

<div class="func"><table class="func" width="100%"><tr><td class="tdleft"><code class="func">‣ BooleanFunctionByNeuralNetworkDASG</code>( <var class="Arg">Func</var> )</td><td class="tdright">( function )</td></tr></table></div>
<p>For the Boolean function <code class="code">Func</code> the function <code class="code">BooleanFunctionByNeuralNetworkDASG</code> returns a two-layered neural network which realizes <code class="code">Func</code> (see <code class="func">NeuralNetwork</code> (<a href="chap4_mj.html#X780C027786A8E34B"><span class="RefLink">4.1-1</span></a>)). The realization of this function is based on decomposition of <code class="code">Func</code> by the non-unate variables with the biggest influence. The DASG algorithm (DASG - Decomposition Algorithm for Synthesis and Generalization) was proposed in <a href="chapBib_mj.html#biBSubirats2008">[SJF08]</a>, however we use a slightly modified version of this algorithm.</p>


<div class="example"><pre>

<span class="GAPprompt">gap></span> <span class="GAPinput">f:=LogicFunction(3,2,[0,0,0,0,0,1,1,0]);</span>
< Boolean function of 3 variables >
<span class="GAPprompt">gap></span> <span class="GAPinput">nn:=BooleanFunctionByNeuralNetworkDASG(f);</span>
< neural network with 2 threshold elements on inner layer and conjunction on outer level >
<span class="GAPprompt">gap></span> <span class="GAPinput">Display(last);</span>
Inner Layer:
[ [[ 1, 4, 2 ], 3], [[ 1, -4, -2 ], -3] ]
Outer Layer: conjunction
Neural Network realizes the function f :
Boolean function of 3 variables.
[ 0, 0, 0 ] || 0
[ 0, 0, 1 ] || 0
[ 0, 1, 0 ] || 0
[ 0, 1, 1 ] || 0
[ 1, 0, 0 ] || 0
[ 1, 0, 1 ] || 1
[ 1, 1, 0 ] || 1
[ 1, 1, 1 ] || 0
Sum of Products:[ 5, 6 ]

</pre></div>


<div class="chlinkprevnextbot"> <a href="chap0_mj.html">[Top of Book]</a>   <a href="chap0_mj.html#contents">[Contents]</a>    <a href="chap3_mj.html">[Previous Chapter]</a>    <a href="chap5_mj.html">[Next Chapter]</a>   </div>


<div class="chlinkbot"><span class="chlink1">Goto Chapter: </span><a href="chap0_mj.html">Top</a>  <a href="chap1_mj.html">1</a>  <a href="chap2_mj.html">2</a>  <a href="chap3_mj.html">3</a>  <a href="chap4_mj.html">4</a>  <a href="chap5_mj.html">5</a>  <a href="chapBib_mj.html">Bib</a>  <a href="chapInd_mj.html">Ind</a>  </div>

<hr />
<p class="foot">generated by <a href="http://www.math.rwth-aachen.de/~Frank.Luebeck/GAPDoc">GAPDoc2HTML</a></p>
</body>
</html>

100%


¤ Dauer der Verarbeitung: 0.16 Sekunden  (vorverarbeitet)  ¤

*© Formatika GbR, Deutschland




Wurzel

Suchen

Beweissystem der NASA

Beweissystem Isabelle

NIST Cobol Testsuite

Cephes Mathematical Library

Wiener Entwicklungsmethode

Haftungshinweis

Die Informationen auf dieser Webseite wurden nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit, noch Qualität der bereit gestellten Informationen zugesichert.

Bemerkung:

Die farbliche Syntaxdarstellung ist noch experimentell.