文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > 高等教育 > 经济学 > reconstructing embedding spaces
侵权投诉

reconstructing embedding spaces

混沌文献

PHYSICALREVIEWE,VOLUME65,035204 R

Reconstructingembeddingspacesofcoupleddynamicalsystemsfrommultivariatedata

S.Boccaletti,1,2D.L.Valladares,2,3LouisM.Pecora,4HiteP.Geffert,4andT.Carroll4

2

IstitutoNazionalediOtticaApplicata,LargoE.Fermi,6,I50125Florence,Italy

DepartamentodeFisicayMatematicaAplicada,UniversidaddeNavarra,Pamplona,Spain

3

DepartamentodeFisica,UniversidadNacionaldeSanLuis,SanLuis,Argentina

4

Code6343,NavalResearchLaboratory,Washington,D.C.20375

Received4October2001;published21February2002

1

Amethodforreconstructingdimensionsofsubspacesforweaklycoupleddynamicalsystemsisoffered.Thetoolisabletoextrapolatethesubspacedimensionsfromthezerocouplinglimit,wherethedivisionofdimensionsasperthealgorithmisexact.Implementationoftheproposedtechniquetomultivariatedatademonstratesitseffectivenessindisentanglingsubspacedimensionalitiesalsointhecaseofemergentsynchro-nizedmotions,forbothnumericalandexperimentalsystems.DOI:10.1103/PhysRevE.65.035204

PACSnumber s :05.45.Tp,05.45.Xt,45.10. b

Thecharacterizationofchaoticandirregularmeasuredtimeserieswiththehelpofnonlinearanalysistoolsisasubjectofgreatscienti cinterest 1 .Apreliminaryfunda-mentaltaskinallanalysistechniquesishowtoreconstructcorrectlythestrangechaoticattractorsfromobservedscalardata.Tothispurpose,ithasbeenshownthatanembeddingcanbeobtainedfromtime-delayedcoordinatesoftheob-servedvariables 2 ,bymeansofwhichonecanreconstructthechaoticattractorofagivendynamics,andthenuseitforcharacterization 3 ,prediction 4 ,measurement,modeling,orcontrol 5 purposes.Alternativeprocedureshavebeenproposedbasedonthereconstructionofthemainfeaturesofthechaoticdynamicsbymeansoftheinterspikeintervaltechnique 6 ,orbyadaptivemethods 7 .TheembeddingtechniqueofRef. 2 dependsonthesuitablechoiceoftwoparameters,namelytheembeddingtimeandtheembeddingdimensions.Whiletheformermaybeeasilyobtainedasthe rstminimumofthemutualinformationfunction 8 ,acor-rectdeterminationofthesystemdimensionalityisanessen-tialproblemtobesolvedinordertoapproachcorrectlyanyfollowingstepsofanonlineardataanalysis.

Apopularmethodthatisusedformeasuringtheminimalembeddingdimensionistheso-calledfalsenearest-neighbor FNN method,originallyintroducedbyKenneletal. 9 ,andlaterimprovedandreelaboratedinordertofacespeci canalysistasks 10 .Themethodconsistsinmarkingasfalsenearestneighborsatdimensionmthosepairsofnearest-neighborm-dimensionalembeddedvectorswhosedistanceatdimensionm 1exceedsagivennumberoftimestheirdis-tanceatdimensionm,thusaccountingforpossibleself-intersectionsofthe owduetoinsuf cientdimensionalityintheembeddedspace.AvanishingfractionofFNNmarkstheminimumdimensionalityneededtoproperlyreconstructthechaotic ow.Thistechniquehasbeenlaterimproved 10 ,alsocomplementingthisanalysiswiththeoneonthesignalsurrogates 11 .

Inthispaper,wediscusshowtoextendthedimensionreconstructionproblemtothecaseofmultivariatedataanalysis,thatis,tothecaseinwhichanobserverispresentedwithasystemcomposedbynweaklycouplednonidenticaldynamicalsubsystems ofdimensionsl1,l2,...,ln,respec-tively ,andextractsseparatelyscalarquantitiesxi(t)outofeachsubsystemi(i 1,...,n).Iftheobserverisinterestedinprobingglobalpropertiesofthesystemunderstudy andifthesubsystemvariablesarealltoallcoupled ,thentheusualreconstructionmethodsworkregardlessoftheparticularvariablexi(t)onwhichembeddingisperformed.However,thisgivesinformationonthefullphase-spacedimensionality.Therearerelevantcases,suchasdetectionofsynchroniza-tionprocesses 12,13 ,ordetectionofinterdependenceofdynamics 14 ,forwhichthedeterminationofthedimen-sionalityofeachsubsystemisneeded.Synchronizationfea-tureshavebeenlargelystudiedforbothcoupledchaotic 12,13 andspace-timechaotic 15 systems,aswellasana-lyzedandstudiedinexperimentsandnaturalphenomena 16 .Recently,variousattemptsatprovidingaunifyingfor-malismencompassingthemajorsynchronizationfeatureshavebeenmade 17 ,whoseapplicationstorealdatarelyonaproperdeterminationofthesubspacedimensionsontowhichtheglobaldynamicsshouldbeprojectedtolookforfunctionalrelationships.

Theproblemcanbestatedasfollows.Considerhavingnweaklycouplednonidenticaldynamicalsystems,andsup-posethatanobserverisabletomeasureseparatelyscalarquantitiesxi(t)outofeachsubsystemi(i 1,...,n).Inordertoproperlyreconstructthedimensionsl1,l2,...,ln,letusconsiderthevector

¯z x1 t ,x1 t 1 ,x1 t 2 1 ,...,x1 t m1 1 1 ,

x2 t ,x2 t 2 ,...,x2 t m2 1 2 ,...,xn t ,xn t n ,...,xn t mn 1 n …,

1

where i(i 1,...,n)arendifferentembeddingtimes.¯zm R(m i 1,nmi)isavectorwhose rst second,third,...,nth)m1(m2,m3,...,mn)componentsresultfromtheembeddingofthex1(x2,x3,...,xn)scalarvariablewithembeddingtime 1( 2, 3,..., n).Theem-beddingtimes icanbedifferentfromeachother,sincedif-ferentobservedvariablesxi(t)mayshowdifferentmutualinformationproperties 8 .

第1页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email