Second order Poincaré inequalities and CLTs on Wiener space
16
pages
English
Documents
Obtenez un accès à la bibliothèque pour le consulter en ligne En savoir plus
Découvre YouScribe et accède à tout notre catalogue !
Découvre YouScribe et accède à tout notre catalogue !
16
pages
English
Documents
Obtenez un accès à la bibliothèque pour le consulter en ligne En savoir plus
Publié par
Langue
English
Second order Poincaré inequalities and CLTs on Wiener space by Ivan Nourdin ? , Giovanni Peccati † and Gesine Reinert ‡ Université Paris VI, Université Paris Ouest and Oxford University Abstract: We prove infinite-dimensional second order Poincaré inequalities on Wiener space, thus closing a circle of ideas linking limit theorems for functionals of Gaussian fields, Stein's method and Malliavin calculus. We provide two applications: (i) to a new second order characterization of CLTs on a fixed Wiener chaos, and (ii) to linear functionals of Gaussian-subordinated fields. Key words: central limit theorems; isonormal Gaussian processes; linear functionals; multi- ple integrals; second order Poincaré inequalities; Stein's method; Wiener chaos 2000 Mathematics Subject Classification: 60F05; 60G15; 60H07 1 Introduction Let N ? N (0, 1) be a standard Gaussian random variable. In its most basic formulation, the Gaussian Poincaré inequality states that, for every di?erentiable function f : R? R, Varf(N) 6 Ef ?(N)2, (1.1) with equality if and only if f is a?ne. The estimate (1.1) is a fundamental tool of stochastic analysis: it implies that, if the random variable f ?(N) has a small L2(?) norm, then f(N) has necessarily small fluctuations.
- dimensional poincaré
- isonormal gaussian process
- standard gaussian
- malliavin
- multi-dimensional
- operator ?
- gaussian subordinated
- random variable
- poincaré inequalities
Publié par
Langue
English
y z
NN (0; 1)
f :R!R
0 2Varf(N)6Ef (N) ;
f
0 2f (N) L ( ) f(N)
X
1;2H F2D
X F DF
H
2VarF 6EkDFk ;H
F
X
y
z
inStein'sPmethoreferencesdifand200Malliaersionvintcalculus.shallWceCacoullosproofvidesometderivwWienerorstapplications:momen(i)UnivtoMoaivnewHousecondedorderrecocisonormalharacterizationSectionoffunctionalCLyTits(1.2)ontaPropxersionebasedddeWienerVI),c05,haos,Pand92000(ii)Email:toChenlinearerez-AbreufunctionalstheofhGaussian-subcalculus)ordfolloiParisnatedcesselds.ertKeyReinertwaordsThen,:GiocenistralinlimitIvtheorems;CLisonormaliGaussianformproelemencesses;oflinearelofunctiaovnevahniqueslLabs;etmMarieulti-88,plePinivan.nourdin@upmc.frtegrals;'XsecondNanorderuePLSToincar?Vinequalities;vStein'sal.metho6,d;andWandiInenerproc[10]heaoMalliasw2000erMathematicsinnite-dimensionalSub(1.1).jecteClassication:p60F05;v60G15;separable60H07Universit?1andInandtrobductionvin-dierenLetvelds,MalliaaneidenotedGaussNourdinofrandomfunctionalsvb,ethabstandardsGaussianlitiesrandomequalitvandariable.hasInaitssmostofbasiccform.ulation,3.1thewGaussianvPgeneraloincar?(1.2),inequalitcenyofstatesordersthat,theforelopevNoteerytdierenrobabtiabled?lesfunctionPierrefor(PtheoremscourrierlimitpllinkingJussieu,ideasris,rance.ofEquipcircleaaUnivclosingOuestuslathvspace,laWienerterre,onUinequalitiesPoincar?FPil.comorder[2],secondetinnite-dimensional,(1.1)[5,with7],equalitdr?yPif[10],andtheonlytherein.ifparticular,eresultsisvane.inThe(whicestimakmuseatethe(1.1)vinisalloatofundamenvtalthetowingolvofofstoLetcbhasticananalysis:Gaussianitroimpolieserthat,realifHilbthespacerandom(seev2),ariableletvGesineproeccatieanniWeAbstract:Malliahastiableaofsmall.UniversitythedvinOxforativnorm,ofthen,andbOuest,Paris,Universit?ahaselemennecessarilywithsmallaluesuctuations.anRelandaholdstiatoynspace(1.1)onhasTbandeeninequarstwithproyvfedonlyboincar?ytheNashofinconstan[14],pluandanthentrediscothevWienereredhaosbPyInChernoositioninb[9]w(beothproproeofsmoreusevHermiteofpinolynomials).olvingThetralGaussiantsParbitraryoincar?eninequalitandyonadmitstecextensionsdevinedsev[16].eralorderdirections,oraencompassingoirebPothilit?stheMocaseAl?atoires,ofersit?smoetVI,CurieDepartmenarisofBo?teUniv1y4Oxford,aSouthearks75252OxfordaX1CedexUK.Freinert@stats.ox.ac.Email:1Secondinnite-dimensional)eGaussiandelds,land,ofersit?non-GaussianarisprobaterrebD?fense,ilitAyendistributideoR?publique,nNansandA,seene.g.ersit?BakryarisetI,al.rance.[1],giovanni.peccati@gmaBobkoothtfunctionalsStatistics,ofersitmofulti-d1imePnRoad,sionalO(a3TG,nEmaildukpossiblyF =f(X ;:::;X )1 d
N (0; 1) X ;:::;X1 d
2VarF 6Ekrf(X ;:::;X )k ;1 d dR
rf f
F =f(X ;:::;X ) X ;:::;X1 d 1 d
N (0; 1) f
dd Hessf
rf
F
E(F ) =
2VarF = > 0 ZN ( ; ) d (F;Z)TV
F Z
p
2 5 1 14 4
4 4d (F;Z)6 E[kHessf(X ;:::;X )k ] E[krf(X ;:::;X )k ] ;TV 1 d 1 d dop R2
kHessf(X ;:::;X )k Hessf(X ;:::;X )1 d op 1 d
dW
X
2;4H F2D
2 2E (F ) = Var (F ) = > 0 ZN ( ; )
p
1h i1 10 442 4 4d (F;Z)6 E kD Fk E kDFk :W op H2
F
p
1h i1 10 42 4 44d (F;Z)6 E kD Fk E kDFk :TV op H2
2;4D
2 2D F H
2 H D F
op
2f7!hf;D FiH
rand(3.21).withseeet,isandacofthewsrandomla)theGaussianeenvwassessingetpapbdieren(1.4)ariableswherewithdistancetheariationy)vetotalhingthethey(seeb,denoteoutandthe,eb,),.tfunctionisoththesymmetricopwneratortlynormspofbtheess(random)ablemandatrix.Leof2.2]).laTheoremb[4,.(seestatewsandfollofoashasesIf,gois.ctAe,relationthatsucthehuppaswhere(1.4)classisMalliacalledSectionaasecondisorderofPeoincar?theinequalitnormyectral:hmidtitconietsanproovsomeedsepinert[4]vbandywithcomumebininGaussiangand(1.3)ofwitheenanLadequate3.2,variationersiontotalofinequalitStein'sonemethoonlydmatrix(seealsoeif.oing[4],.recen[addition8of,c24]).espInthe[16,meRemarkIn3.6]tthsuceagainrstariabletthatresultwThenttheThethenro(1.6)ofi.i.d.1.1tdetailedtiableSectiondenedAsnoteina4.2,elemencrucialaluesoint,isproTheoremwithleadsthatfurtherthvelluseful)towhiceratowequivnametheconof2Hilb(1.4)eratorareasspthatecialLinstancesariance.ofemisonormalucprhcmoreovergeneralrestimates,alwhicarhHilbcanspbee,obtainedletbmeanymatccomariablebiningAssSthattein'srandommethoadwandlaMalliaandvinwcalculustheonwan.innite-dimensionaletGaussian(3.21))space.SectionItdistanceisvthereforeThennaturalthetoyaskanwhethercanthethenresultsonofnot[16],canHessianbonecusesusedoneinthatotedrdperChatterjeetoerobtainta(1.5)generalinv,ersionlawofthe(1.4),absolutelyinontinuousvrolvingeatodistanceLtoesgueGaussianasurforthensmotiable.othwicefunctionalsisofharbitraryisinnite-dimensionali.i.d.Gaussianaree(wherelvds.theWoseesshallNoshoofwgradienthatisthe(1.3)answ,ervisTheprandomositivofe.ofIndeed,wiceonvin-dierenefunctionalsofformallytheinprin2;cipalthatacsmohievisemenrandomtstofvthisinpapifer(theistensortheductproaofitselfofandthewfollousedwingfactstatemen-knotw(casepreciseindicateTheopdenotesrthe(or,Walenasserstein,distance,spseeradius)(3.22)):theTheoremom1.1ert-Sc(Secondoporderecialinanite-dimensionaltainsP(1.2)oincar?inequalitw.opauthorsofofTheoremtheispresenint4.1.papdiscussederSectionpaoinptedtoutthatthat1.1thetonite-(anddimensionaleryStein-tinequalities,ypheeinequalitiesrandomleadingtractiontoRelation2D F
F = (F ;:::;F )1 d
qH q> 1 H q
qH H q X =fX(h);h2
Hg H
( ;F;P ) X
H E [X(h)X(g)] = hh;gi FH
X
q> 1 H q Xq
2L ( ;F;P ) fH (X(h));h2H;khk = 1gq H
2 2x q xq d
2 2H q H (x) = ( 1) e eq q qdx
qH =R q > 1 I (h ) =q!H (X(h))0 q q
qHp
q!k k q H q = 0 I (c) =c c2R
q q 0H
2L ( ;F;P )
Hq
2F2L ( ;F;P )
1X
F = I (f );q q
q=0
qf = E[F ] f 2H q > 1 F q > 00 q
J qq
2F2L ( ;F;P ) J F =I (f ) q> 0q q q
p qfe ; k > 1g H f2 H g2 Hk
(p+q 2r)r = 0;:::;p^q f g r H
,noonwrandpresensumtrecenth4einbasicexpansionelemenotstainedof,GaussiantheanalysisFandeMalliaasvinwherecalculusalsothatcenare23],used2in.this1.1,paptoer.equippTheypreaderSectioniswnreferredetoanthTheestandwboWienermonographsInb(2.7),yeryMalliaorthonormalvinresult[12]tandaNualarty[19]asforproancanybunexplainedductdenitionmooinequalitiesrhaosresult.3LetMallianecessaryItbexpansion)eInatorealwsseparabletegrableHilbpapertfollospace.ofFeorexploredanoy,newFofeofproletthpro(CLtheStein'stoMalliabforeatheb5)inthuniestensorforproductcofenSectioneandordenoteconbmappingyw(seeofleadwithwillprobabilities.theeassolinearciatedwandtensorthetsymmetricwithtenormnbsorandproWienerduoincar?ct.PWeeopwritev,preliminaryewvctisomea2derivcansecondosedtheinniteftheo.ersionsquarevvtoisindicateofanadmitsisonormalcGaussubsianlinearproCLcess(2.7)oSectionviserinequalitiestractedtrac,roleduniquelye[15]).nevedspaceonbsometheprobabilitopythespcacifelimitcondaasecalculus,olvideasvcircleinery..Thisvirtuallymea20,nasithatGivestimatesndingsisandaariable.ceneryteredandGaussianefamily,,actionwhosestandardcotionvquarianceiniFsangivtractionenrandomithenwithtermsellofastheTheoreminnerofprotheductdealsofSectionTheseeenbbyextended.ainequalitiesisometryinetergeseenvsymmetricconprohaoswcbWieneredxedthetodiedelongingdistancesboundsariablesandverandomthe.thWcetalso.assumeorthatconcernsofwiswritegeneratederators.bvinyolvingsequence,.resultsF.orisevell-knoery(Wienertohaoswthatlarecallhw3Sectionensure.thatbbdecompeinthetheaorthogonalthofWienerspacescfollohaosTherefore,ofyein,randomthatariableis,organizedtheerclosedthelinearrestsubspelds.athecewingofhaoticypordinatedtGaussianthefunctionalsofforariablesTvstudyrandomwforwheregenerated6,binyfurtherthe,randomthevnariablestiofcontheoftTheypareedetermined(1.5)yof.ersionorverya(seetowoteddenotedevyison7orthogonalSectionjection,eratorFinallynelds.s)ordinatedWienersubhaos.Gaussianparticular,forTstheoremsTtral,andwheremethoCLisninisthenthevinolinkingthtHermiteofptfulolynomialevdenedfruiasvandLethaosclosescandWiener21,on[16,seTcompleteCLsystemonn,.elyenectivconrespthecus,andfogeneralizes6ThisSection,andev5PreliminariesSection.ities.suWciofnshallconditionsthethetontreofwhicandmlorder.isWelemeneofwritevboyGaussianconav,let1X
f
g = hf;e
:::
e i
r
hg;e
:::
e i
r:r i i i i1 r H 1 r H
i ;:::;i =11 r
ef
g f
g2r r
