摘要
Thispaperpresentsanewtypeofcellularautomata(CA)modelforthesimulationofalternativelanddevelopmentusingneuralnetworksforurbanplanning.CAmodelscanberegardedasaplanningtoolbecausetheycangeneratealternativeurbangrowth.AlternativedevelopmentpatternscanbeformedbyusingdifferentsetsofparametervaluesinCAsimulation.Acriticalissueishowtodefineparametervaluesforrealisticandidealizedsimulation.ThispaperdemonstratesthatneuralnetworkscansimplifyCAmodelsbutgeneratemoreplausibleresults.Thesimulationisbasedonasimplethree-layernetworkwithanoutputneurontogenerateconversionprobability.Notransitionrulesarerequiredforthesimulation.Parametervaluesareautomaticallyobtainedfromthetrainingofnetworkbyusingsatelliteremotesensingdata.Originaltrainingdatacanbeassessedandmodifiedaccordingtoplanningobjectives.Alternativeurbanpatternscanbeeasilyformulatedbyusingthemodifiedtrainingdatasetsratherthanchangingthemodel.
出版日期
2004年01月11日（中国期刊网平台首次上网日期，不代表论文的发表时间）
