BP神经网络结合遗传算法寻优过程.zip_neural network_BP神经网络

<html xmlns="http://www.w3.org/1999/xhtml"> <head> <meta charset="utf-8"> <meta name="generator" content="pdf2htmlEX"> <meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1"> <link rel="stylesheet" href="https://static.pudn.com/base/css/base.min.css"> <link rel="stylesheet" href="https://static.pudn.com/base/css/fancy.min.css"> <link rel="stylesheet" href="https://static.pudn.com/prod/directory_preview_static/626e51c515314c70c394ed25/raw.css"> <script src="https://static.pudn.com/base/js/compatibility.min.js"></script> <script src="https://static.pudn.com/base/js/pdf2htmlEX.min.js"></script> <script> try{ pdf2htmlEX.defaultViewer = new pdf2htmlEX.Viewer({}); }catch(e){} </script> <title></title> </head> <body> <div id="sidebar" style="display: none"> <div id="outline"> </div> </div> <div id="pf1" class="pf w0 h0" data-page-no="1"><div class="pc pc1 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/626e51c515314c70c394ed25/bg1.jpg"><div class="c x0 y1 w2 h2"><div class="t m0 x1 h3 y2 ff1 fs0 fc0 sc0 ls0 ws0">BP 神经网络的工程应用及其基于遗传算法寻优过程</div><div class="t m0 x2 h4 y3 ff2 fs1 fc0 sc1 ls0 ws0">摘要：BP 神经网络对数据复杂的映射关系（比如非线性）具有良好的拟合性能，</div><div class="t m0 x2 h4 y4 ff2 fs1 fc0 sc0 ls0 ws0">合适的网络结构可以无限逼近拟合函数。遗传算法是基于生物进化的思想，对</div><div class="t m0 x2 h4 y5 ff2 fs1 fc0 sc0 ls0 ws0">解域进行遗传运算找到最优个体（最优解），及其对应的最优适应度值（最优</div><div class="t m0 x2 h4 y6 ff2 fs1 fc0 sc0 ls0 ws0">值）。</div><div class="t m0 x2 h4 y7 ff2 fs1 fc0 sc1 ls0 ws0">关键字：BP 神经网络，遗传算法，极值寻优</div><div class="t m0 x2 h4 y8 ff4 fs1 fc0 sc0 ls0 ws0">1. 引论</div><div class="t m0 x2 h4 y9 ff3 fs1 fc0 sc0 ls0 ws0">BP 神经网络是工程应用最广的网络之一。工程应用领域有非线性函数拟合、模</div><div class="t m0 x2 h4 ya ff2 fs1 fc0 sc0 ls0 ws0">式识别等。BP 网络可以看作“黑匣子”用于拟合输入输出间非线性关系，合理结</div><div class="t m0 x2 h4 yb ff2 fs1 fc0 sc0 ls0 ws0">构的 BP 网络能够无限逼近期望输出。拟合性能较好的 BP 网络可以应用遗传算</div><div class="t m0 x2 h4 yc ff2 fs1 fc0 sc0 ls0 ws0">法找到最优的网络输出及其对应输入值。</div><div class="t m0 x2 h4 yd ff4 fs1 fc0 sc0 ls0 ws0">2. BP 神经网络概述</div><div class="t m0 x2 h4 ye ff3 fs1 fc0 sc0 ls0 ws0">BP 网络是一种多层前馈神经网络，其主要特点是信号前向传递，误差反向传递。</div><div class="t m0 x2 h4 yf ff2 fs1 fc0 sc0 ls0 ws0">在前向传递中，输入信号从输入层经隐藏层逐层处理，直至输出层。每一层神</div><div class="t m0 x2 h4 y10 ff2 fs1 fc0 sc0 ls0 ws0">经元状态值影响下一层神经元状态。如果输出层得不到期望输出，则转入反向</div><div class="t m0 x2 h4 y11 ff2 fs1 fc0 sc0 ls0 ws0">传播，根据预测误差调整权值和阀值，从而使 BP 神经网络输出不断接近期望</div><div class="t m0 x2 h4 y12 ff2 fs1 fc0 sc0 ls0 ws0">输出。</div><div class="t m0 x2 h4 y13 ff3 fs1 fc0 sc0 ls0 ws0">BP 网络预测首先要训练网络，通过训练使网络具有联想记忆和预测能力。训练</div><div class="t m0 x2 h4 y14 ff2 fs1 fc0 sc0 ls0 ws0">步骤有：</div><div class="t m0 x2 h4 y15 ff2 fs1 fc0 sc0 ls0 ws0">步骤一：网络参数初始化。</div><div class="t m0 x2 h4 y16 ff2 fs1 fc0 sc0 ls0 ws0">步骤二：隐含层输出计算。</div><div class="t m0 x2 h4 y17 ff2 fs1 fc0 sc0 ls0 ws0">步骤三：输出层输出计算。</div><div class="t m0 x2 h4 y18 ff2 fs1 fc0 sc0 ls0 ws0">步骤四：误差计算。</div><div class="t m0 x2 h4 y19 ff2 fs1 fc0 sc0 ls0 ws0">步骤五：权值更新。</div><div class="t m0 x2 h4 y1a ff2 fs1 fc0 sc0 ls0 ws0">步骤六：阀值更新。</div><div class="t m0 x2 h4 y1b ff2 fs1 fc0 sc0 ls0 ws0">步骤七：迭代终止判断，若没有结束，返回步骤二。</div></div></div><div class="pi" data-data='{"ctm":[1.611850,0.000000,0.000000,1.611850,0.000000,0.000000]}'></div></div> </body> </html>

<div id="pf2" class="pf w0 h0" data-page-no="2"><div class="pc pc2 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/626e51c515314c70c394ed25/bg2.jpg"><div class="c x0 y1 w2 h2"><div class="t m0 x2 h4 y1c ff4 fs1 fc0 sc0 ls0 ws0">3. 遗传算法概述</div><div class="t m0 x2 h4 y1d ff2 fs1 fc0 sc0 ls0 ws0">遗传算法（Genetic Algorithms）是 1962 年有美国 Michigan 大学 Holland 教授提</div><div class="t m0 x2 h4 y1e ff2 fs1 fc0 sc0 ls0 ws0">出的模拟自然界遗传机制和生物进化论而成的一种并行随机搜索最优化方法。</div><div class="t m0 x2 h4 y1f ff2 fs1 fc0 sc0 ls0 ws0">把自然界进化原理引入优化参数行程的编码串联群体中，按照所选择的适应度</div><div class="t m0 x2 h4 y20 ff2 fs1 fc0 sc0 ls0 ws0">差的个体被淘汰，新的群体继承了上一代的信息，又优于上一代。这样反复循</div><div class="t m0 x2 h4 y21 ff2 fs1 fc0 sc0 ls0 ws0">环，直至满足条件。算法基本操作流程如下图。</div><div class="t m0 x3 h5 y22 ff2 fs2 fc0 sc0 ls0 ws0">开始</div><div class="t m0 x4 h5 y23 ff2 fs2 fc0 sc0 ls0 ws0">产生初始种群</div><div class="t m0 x4 h5 y24 ff2 fs2 fc0 sc0 ls0 ws0">判断停止条件</div><div class="t m0 x5 h5 y25 ff2 fs2 fc0 sc0 ls0 ws0">计算适应值函数</div><div class="t m0 x6 h5 y26 ff2 fs2 fc0 sc0 ls0 ws0">选择</div><div class="t m0 x7 h5 y27 ff2 fs2 fc0 sc0 ls0 ws0">遗传运算</div><div class="t m0 x8 h5 y28 ff2 fs2 fc0 sc0 ls0 ws0">更新种群</div><div class="t m0 x9 h4 y29 ff2 fs1 fc0 sc0 ls0 ws0">输出</div><div class="t m0 xa h4 y2a ff2 fs1 fc0 sc0 ls0 ws0">停止</div><div class="t m0 xb h6 y2b ff3 fs1 fc0 sc0 ls0 ws0">Y</div><div class="t m0 xc h6 y2c ff3 fs1 fc0 sc0 ls0 ws0">N</div></div></div><div class="pi" data-data='{"ctm":[1.611850,0.000000,0.000000,1.611850,0.000000,0.000000]}'></div></div>