Web-Books
in the Austria-Forum
Austria-Forum
Web-Books
Naturwissenschaften
Physik
Differential Geometrical Theory of Statistics
Page - 193 -
  • User
  • Version
    • full version
    • text only version
  • Language
    • Deutsch - German
    • English

Page - 193 - in Differential Geometrical Theory of Statistics

Image of the Page - 193 -

Image of the Page - 193 - in Differential Geometrical Theory of Statistics

Text of the Page - 193 -

Entropy2016,18, 433 Proposition9. Weassumeamodel (Θ,P) is regular. For everynonzero realnumberαonehas M(āˆ‡Ī±,āˆ‡āˆ’Ī±)∩M(āˆ‡āˆ’Ī±,āˆ‡Ī±)=M(g,āˆ‡āˆ’Ī±,āˆ‡Ī±)+M(g,āˆ‡āˆ—,āˆ‡). TheSketchofProof. Theproof isbasedontheshortexact sequences 0→M(g,āˆ‡Ī±,āˆ‡āˆ’Ī±)→M(āˆ‡Ī±,āˆ‡āˆ’Ī±)→Sāˆ‡2 (Θ)→0, 0→M(g,āˆ‡āˆ’Ī±,āˆ‡Ī±)→M(āˆ‡āˆ’Ī±,āˆ‡Ī±)→Sāˆ’āˆ‡2 (Θ)→0. Letussuppose that theconclusionof the theproposition fails. Thenthere isanonzero2-form B∈Sāˆ‡Ī±2 (Θ)∩Sāˆ‡ āˆ’Ī± 2 (Θ). (1) IfKer(B)=0, thenbothāˆ‡Ī± andāˆ‡āˆ’Ī± coincidewith theLevi-CivitaconnectionofB. This implies α=0, thiscontradictsourchoiceofα. (2) If Ker(B) = 0 then Ker(B) and Ker(B)⊄ are geodesic for bothāˆ‡Ī± andāˆ‡āˆ’Ī±. Thus the pair (Ker(B),Ker(B)⊄)definesag-orthogonal2-web. Atoneside,by thevirtueof the reduction theoremas in [18]every leafFofKer(B)⊄ inheritsa dualpair (F,gF,āˆ‡Ī±F,āˆ‡āˆ’Ī±F ). Atanotherside,Bgivesrise to theRiemannianstructure (F,B). Furthermorebothāˆ‡Ī±F andāˆ‡āˆ’Ī±F are torsionfreemetricconnections in (F,B). Therebyonegets āˆ‡Ī±F=āˆ‡āˆ’Ī±F The last equalityholds if andonly ifα= 0.This contradictsourassumption. Theproposition isproved. Thepropositionabove isaseparationcriterionforα-connections in the followingsense. Forevery nonzerorealnumberα thevectorsubspaceSāˆ‡Ī±2 (Θ) is transverse toSāˆ‡ āˆ’Ī± 2 (Θ) in thevectorspaceS2(Θ) ofsymmetric forms inΘ. 5.4. TheExponentialModels and theHyperbolicity Achallengeisthesearchofacriterionfordecidingwhetheramodel(Θ,P) isanexponentialfamily. That is thechallenge in [22]. Still,nowadays, thisproblemisopen. TheFisher informationofaregularexponentialmodel isaHessianRiemannianmetric.Weare goingtodemonstratethat theconverseisgloballytrue.Ourproof isbasedoncohomologicalarguments. In theAppendixAtothispaperweintroduceanewnumerical invariantrb(Θ,P)whichmeasures howfar frombeinganexponential family isamodel (Θ,P). The invariant rbderives fromtheglobalanalysisofdifferentialoperators Dα=Dāˆ‡Ī±, Dα=Dāˆ‡ α . Nowwearegoingtoprovideacohomological characterizationofexponentialmodels. Beforepursuingwerecalladefinition. LetĪøj, j :=1,...,mbeasystemofEuclideancoordinate functionsofRm. Definition40. [18]Anm-dimensional statisticalmodel (Θ,P) is called an exponentialmodel for (Īž,Ī©) if there exist amap Īž ξ→ [C(ξ),F1(ξ),...,Fm(ξ)]∈Rm+1 193
back to the  book Differential Geometrical Theory of Statistics"
Differential Geometrical Theory of Statistics
Title
Differential Geometrical Theory of Statistics
Authors
FrƩdƩric Barbaresco
Frank Nielsen
Editor
MDPI
Location
Basel
Date
2017
Language
English
License
CC BY-NC-ND 4.0
ISBN
978-3-03842-425-3
Size
17.0 x 24.4 cm
Pages
476
Keywords
Entropy, Coding Theory, Maximum entropy, Information geometry, Computational Information Geometry, Hessian Geometry, Divergence Geometry, Information topology, Cohomology, Shape Space, Statistical physics, Thermodynamics
Categories
Naturwissenschaften Physik
Web-Books
Library
Privacy
Imprint
Austria-Forum
Austria-Forum
Web-Books
Differential Geometrical Theory of Statistics