基于人工神经网络的农业生态环境信息分析模型研究1

分类号：论文的分类号UDC：D10621-347-(2011)0501-0密级：公开编号：学号成都信息工程学院硕士学位论文论文题目：基于人工神经网络的农业生态环境信息分析模型研究姓名学号学院学位类型□学术型□专业学位门类(类别)农业推广硕士专业(领域)农业资源利用研究方向导师校内姓名职称校外姓名职称答辩时间年月日基于人工神经网络的农业生态环境信息分析模型研究导师：学生：摘要农业生态系统本身存在一定的不确定性和模糊性，如果用传统的方法模拟这些系统将很困难。神经网络模型作为一种新型模型系统，能较准确地模拟这些系统，因此引起生态学者们的广泛关注。本文系统的研究了误差逆传神经网络模型的算法、结构，并且着重研究了其在生态学和农业领域中的应用问题。采用三层神经网络模型结构的误差逆传神经网络模型，能够模拟较高复杂程度的连续性函数，而且依靠其小巧的结构而很难出现与训练数据相吻合的情况。误差逆传神经网络模型算法具有利用输入误差对权值进行调整的特性。在农业和生态学领域的研究中，预测生物生产量、作物产量、生物与环境之间的非线性函数模拟关系主要依靠误差逆传神经网络模型。随着研究的不断深入，研究人员通过强制训练停止以及复合模型等多种先进技术来提高误差逆传神经网络模型的外推能力，同时对误差逆传神经网络模型的机理的解释提出了敏感性分析法、Garson算法以及随机化检验等。误差逆传神经网络模型可以帮助人们了解模糊性和不确定性较大系统的行为，这成为其较大的技术优势，这些技术优势是传统模型所无法比拟的，因而其可以作为对传统机理模型的重要补充。关键词：人工神经网络，误差逆传，农业及生态系统机理模型AbstractAgro-ecologicalsystemitselfthereisacertainuncertaintyandambiguity,ifthetraditionalmethodtosimulatethesesystemswillbedifficult.Neuralnetworkmodelasanewmodelsystemthatmoreaccuratelysimulatethesesystems,ecologistshavecausedwidespreadconcern.Thissystematicstudyoftheerrorback-passalgorithmforneuralnetworkmodel,structure,anditsfocusonresearchinecologyandagricultureintheapplication.Three-layerneuralnetworkmodelstructureoftheerrorback-passneuralnetworkmodel,tosimulateahighdegreeofcontinuityofcomplexfunctions,butalsoonitsstructureisdifficulttocompactthedataappearconsistentwiththetrainingsituation.Errorback-passalgorithmforneuralnetworkmodelwiththeuseofinputerrorsontherighttoadjustthevalueoftheproperty.Inresearchinthefieldofagricultureandecology,thepredictionofbiologicalproduction,cropproduction,biologicalandenvironmentalnon-linearfunctionmodelingtherelationshipbetweentheerrormainlydependsonretrogradeneuralnetworkmodel.Withthedeepeningofthestudy,theresearchersstoppedbythemandatorytrainingandcompositemodelsandotheradvancedtechnologiestoimprovetheneuralnetworkmodelerrorretrogradeextrapolationcapability,whiletheerrorback-passneuralnetworkmodelproposedtoexplainthemechanismofsensitivityanalysis,Garsonalgorithmsandrandomizationinspection.Errorretrogradeneuralnetworkmodelcanhelppeopleunderstandtheambiguityanduncertaintybehavioroflargersystems,whichbecomelargertechnicaladvantages,thesetechnicaladvantagesunmatchedbythetraditionalmodel,andthereforeitsmechanismasthetraditionalmodelanimportantsupplement.Keywords：Artificialneuralnetworks，Back-propagation，Agro-ecologicalsystem目录摘要..............................................................................................................................2ABSTRACT...................................................................................................................3第一章绪论..................................................................................................................51.1研究的目的和意义............................................................................................51.2文献综述............................................................................................................51.3本课题的研究技术路线及主要内容.................................................................8第二章人工神经网的基本原理..................................................................................92.1人工神经网络的概况........................................................................................92.1.1人工神经网络的概念..................................................................................92.1.2人工神经网络的特点..................................................................................92.1.3人工神经网络的应用与发展......................................................................92.2BP网络模型.....................................................................................................102.2.1BP网络的概念及其基本模型和特征......................................................102.2.2BP网络拓扑结构.....................................................................................112.2.3BP网络的工作原理及过程......................................................................112.2.4BP算法流程..............................................................................................122.3BPN网络模型...................................................................................................132.3.1BPN网络模型的算法和结构....................................................................132.3.2BPN网络模型的检验................................................................................14第三章BPN网络模型在农业及生态学研究中的应用.............................................163.1BPN网络模型的模拟能力...............................................................................163.2BPN网络模型机制的解释...............................................................................163.3BPN网络模型的主要应用范围.......................................................................193.4BPN网络的主要优势和缺陷...........................................................................203.4提高BPN网络模型外推能力的方法..............................................................21第四章农业生态环境信息分析模型........................................................................234.1BP人工神经网络层数的确定........................................................................234.2传递函数和训练参数的确定.............................................................