📄 glossary.properties.fitness.html
字号:
<html dir="LTR">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=Windows-1252" />
<meta name="vs_targetSchema" content="http://schemas.microsoft.com/intellisense/ie5" />
<title>fitness Term</title>
<xml>
</xml>
<link rel="stylesheet" type="text/css" href="MSDN.css" />
<link rel="shortcut icon" href="favicon.ico" type="image/x-icon" />
<link rel="stylesheet" type="text/css" href="DesignNotes.css" />
<script src="DesignNotes.js" language="javascript" type="text/javascript"></script>
</head>
<body id="bodyID" class="dtBODY">
<div id="nsbanner"> <table width="100%" border="0" cellspacing="0" cellpadding="0">
<tr>
<td>
<div id="bannerrow1">
<table class="bannerparthead" cellspacing="0">
<tr id="hdr">
<td class="runninghead" style="font-size: 70%"><a href="javascript:toggle('design_span',true)" title="Show or hide design notes" id="design_toggle" class="show"><img src="D.png" align="absmiddle" onLoad="javascript:check(); focus();"/></a>
Glossary of Genetic Programming Terms</td>
</tr>
</table>
</div>
<div id="TitleRow">
<h1 class="dtH1" style="font-size: 85%; position: relative; left: 5px;">fitness Term</h1>
</div>
</td>
<td align="right" valign="top"><img src="GPE_blue.png"></td>
</tr>
</table></div>
<div id="nstext">
<p>The numeric rank of an <a href="Glossary.Objects.Individual.html">Individual</a> in an <a href="Glossary.Objects.Environment.html">Environment</a>.</p>
<H4 class="dtH4">Remarks</H4>
<P><strong>Fitness</strong> provides a numeric representation of an <a href="Glossary.Objects.Individual.html">Individual</a>'s performance in an <a href="Glossary.Objects.Environment.html">Environment</a>. This is one of the most important parts of the <a href="Glossary.GeneticProgramming.html">Genetic Programming</a> algorithm since it determines how new <strong>Individual</strong>s will be created. Because of this, it is important to choose a <strong>fitness</strong> calculation that matches the goal of the problem.</P>
<P>For example, in the case of the path-finding <em>Artificial Ant</em> problem, the goal is to find all of the food in the minimum number of steps. If you were to measure the only the number of steps taken, you would probably evolve an ant that never moved (the minimum number of steps possible), which would not be very useful.</P>
<P>The idea of <strong>fitness</strong> includes several related concepts [Note: only the common <strong>fitness</strong> processing concepts are listed below]: </P>
<ul>
<li><strong>raw fitness -</strong> an <strong>Individual</strong>'s score from a single <strong>Environment</strong></li>
<li><strong>aggregate fitness -</strong> combines multiple <strong>raw fitness</strong> values from interactions with one or more <strong>Environment</strong>s </li>
<li><strong>normalized fitness -</strong> converts the <strong>aggregate fitness</strong> value to one that is directly comparable to other <strong>normalized fitness</strong> values </li>
</ul>
<p>Sometimes, other metrics (such as the size of an <strong>Individual</strong> or the number of steps that <strong>Individual</strong> took in the environment) are incorporated into the final <strong>fitness </strong>value. </p>
<span id="design_span" class="show">
<div class="design_note">
<h4 style="position: relative; top: -5px;">Design Note</h4><design>
<p>In this framework, the various <strong>fitness</strong> concepts are used as follows: </p>
<ul>
<li><a href="GeneticProgrammingEngine.ProblemSpace.IHistory.RawFitness.html">RawFitness</a> is provided as a property of <a href="GeneticProgrammingEngine.ProblemSpace.IHistory.html">IHistory</a>, as this light-weight recorder of an <strong>Individual</strong>'s actions already knows how the <strong>Individual </strong>has performed. </li>
<li><strong>Fitness</strong>es are aggregated using the <a href="GeneticProgrammingEngine.Internal.EngineOptions.Aggregator.html">Aggregator</a> delegate in the <a href="GeneticProgrammingEngine.Internal.EngineOptions.html">EngineOptions</a> instance passed to the <a href="GeneticProgrammingEngine.GPEngine.Options.html">Options</a> property of the <a href="GeneticProgrammingEngine.GPEngine.html">GPEngine</a>. The default method performs an average of the raw <strong>fitness</strong> values. </li>
<li><strong>Fitness</strong>es are normalized using the <a href="GeneticProgrammingEngine.Internal.EngineOptions.Normalizer.html">Normalizer</a> delegate in the <strong>EngineOptions</strong> instance passed to the <strong>Options</strong> property of the <strong>GPEngine</strong>. The default method simply returns the aggregated <strong>fitness</strong> value. </li>
</ul>
<p>No restrictions are placed on the output values of the <strong>Aggregator</strong> and <strong>Normalizer</strong> methods by the framework, though some <a href="Glossary.GeneticOperations.selection.html">selection</a> techniques require such a format (such as fitness-proportional <strong>selection</strong> requiring a normalized value between 0 and 1). </p>
<p>As for the use of external metrics in the <strong>fitness</strong> calculation, in practice this is problematic. The metric values must be normalized to correspond with the <strong>fitness</strong> values (too large and the only thing you measure is the metric; too small and the metric has no effect) and then combined with one of the <strong>fitness</strong> processing steps (Aggregate it as another fitness value? Use it as a normalization factor?).</p>
<p>Adding such a facility to incorporate arbitrary metrics would have provided limited functionality at the cost of a much more complicated interface, so we decided against it. However, there is nothing to prevent users from incorporating such metrics into their own <strong>fitness</strong> calculations. </p>
</design></div>
</span>
<h4 class="dtH4">See Also</h4>
<p>
<a href="GeneticProgrammingEngine.ProblemSpace.IHistory.html">IHistory Interface</a> | <a href="GeneticProgrammingEngine.ProblemSpace.html">GeneticProgrammingEngine.ProblemSpace Namespace</a> | <a href="Glossary.Objects.History.html">History Term</a> | <a href="GeneticProgrammingEngine.Internal.EngineOptions.html">EngineOptions Class</a> | <a href="GeneticProgrammingEngine.Internal.EngineOptions.MaximizeFitness.html">MaximizeFitness Property</a> | <a href="GeneticProgrammingEngine.Internal.EngineOptions.Aggregator.html">Aggregator Property</a> | <a href="GeneticProgrammingEngine.Internal.EngineOptions.Normalizer.html">Normalizer Property</a></p>
<hr />
<div id="footer"><div align="left"><a href="http://www.opensource.org/" target="_top"><img src="http://opensource.org/trademarks/open_source_button.png" border="0" align="left" style="padding-bottom: 5px"/></a> This software licensed under the <a href="http://opensource.org/licenses/afl-2.1.php" target="_top">Academic Free License version 2.1</a>. (<a href="afl-2.1.txt">Plain text</a>)</div>
<div align="right">Hosting for <a href="http://gpe.sourceforge.net/">this project</a> provided by <a href="http://sourceforge.net/">SourceForge.net</a>. (<a href="http://sourceforge.net/projects/gpe">SourceForge home</a>) <a href="http://sourceforge.net"><img src="http://sourceforge.net/sflogo.php?group_id=140347&type=1" width="88" height="31" border="0" align="absbottom" alt="SourceForge.net Logo" /></a></div>
</div>
</div>
</body>
</html>⌨️ 快捷键说明
复制代码
Ctrl + C
搜索代码
Ctrl + F
全屏模式
F11
切换主题
Ctrl + Shift + D
显示快捷键
?
增大字号
Ctrl + =
减小字号
Ctrl + -