人工神經(jīng)網(wǎng)絡(luò)隱層神經(jīng)元數(shù)的確定.pdf

1、重慶大學(xué)碩士學(xué)位論文人工神經(jīng)網(wǎng)絡(luò)隱層神經(jīng)元數(shù)的確定姓名：王立威申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：應(yīng)用數(shù)學(xué)指導(dǎo)教師：劉瓊蓀20120526重厭大學(xué)碩士學(xué)位論文一——莖蘭壟鐾—————————————————————————————————————————————————————————————————————————一一ABSTRACTAn飾cialneuralnetworkisalloptimizationmodelwhichcanabstrac

2、tanddescriptthehumanbrainneuralnetu，orkfrommathematics，Ph3，sicalandbiologicalaspectsBecauseofabilityofpowerfulmapping，itiswidelyusedinvariousapplicationfieldsHoweverduringtheestablisbanentoftheartificialneuralnetworkmode

3、l，oneofthedifficultiesishowtodeterminethenumberofhidden1ayer—neuronsFirstl、。thepapersystematicallyintroducestheartificialneuralnetworkmodel，andsummarizestheneuralnetworkmodelofhiddenlayerneuronnunlberdeterminationmethodi

4、ndomesticandforeignSecondlyitisbrieflydiscussedonsomecreatl、7emethodsandanalyzedthefactorsofthenumbersofhiddenlayerneuronsThirdly，byintroducingthepolynomialandthethoughtofmatrixpseudoinverse：thepaperwoulddiscussthenumber

5、ofhiddenlayerneuronswhichareneededwhentheneuralnetworkmodelsareestablishedinonedimensionandmulti—dimensionalspacerespectivelyInonedimensionspace，thepaperwouldusethepowerfunctionashiddenlayerneuronactivationfuncti。nandcom

6、binewiththethoughtofmatrixinversetofirstlydetenninetheneuronsoftheoutputweightsandestablishthepolynomialneuralnetworkmodelsFinally，Throughtheintroductionofintervalbinarysearchmethod，thebestnumberofthehiddenla3，erneuronsw

7、ouldbedeterminedInthemulti—dimensionalspacethepaperusesthespecificwa)7todeterminetheinputweightssothatthepurposeofdimensionconversionwou]dberealized，andusestheSfunctionaStheneuronactivationfunctionjcombineswiththethought

8、ofmatrixinverseandfirstlydeterminestheneuronsoftheoutputweights，thepoly7nomialneuralnetworkmodelwouldbeestablishedatIastAndaccordingtothedimensionofthespaceofpolynomialsthenumberofthehiddenlayerneuronsoftheneuralnetworkm