Seite - 79 - in Differential Geometrical Theory of Statistics
Bild der Seite - 79 -
Text der Seite - 79 -
Entropy2016,18, 386
Letωbean invariantvolumeelementonG/K inanaffine local coordinatesystem { x1,x2,...,xn }
inaneighborhoodofo:
ω=Φ ·dx1∧ ...∧dxn (128)
WecanwriteX∗=∑
i χi ∂
∂xi anddeveloptheLiederivativeof thevolumeelementω:
LX∗ω=(LX∗Φ) .dx1∧ ...∧dxn+∑
j Φ.dx1∧···∧LX∗dxj∧···∧dxn= (
X∗Φ+ (
∑
j ∂χj
∂xj )
Φ )
dx1∧ ...∧dxn (129)
Since thevolumeelementω is invariantbyG:
LX∗ω=0⇒X∗Φ+ (
∑
j ∂χj
∂xj )
Φ=0⇒X∗logΦ=−∑
j ∂χj
∂xj (130)
ByusingAX∗Y∗=−DY∗X∗,wehave:(
D ∂
∂xi (AX∗) )(
∂
∂xj )
=D ∂
∂xi (
AX∗ (
∂
∂xj ))
−AX∗ (
D ∂
∂xi ∂
∂xj )
=−D ∂
∂xi D ∂
∂xj (
∑
k χk ∂
∂xk )
=−∑
k ∂2χk
∂xi∂xj ∂
∂xk (131)
ButasD is locallyflatandX∗ isan infinitesimalaffinetransformationwithrespect toD:
D ∂
∂xi (AX∗)=0⇒ ∂ 2χk
∂xi∂xj =0 (132)
TheKoszul formandcanonicalbilinear formaregivenby:
α=∑
i ∂logΦ
∂xi dxi=DlogΦ (133)
Dα=∑
i,j ∂2logΦ
∂xi∂xj dxidxj=DdlogΦ (134)
LX∗α=LX∗DlogΦ=DLX∗logΦ=DX∗logΦ=−D (
∑
j ∂χj
∂xj )
=−∑
,j ∂2χj
∂xi∂xj dxi=0 (135)
Then,LX∗α=0∀X∈ g.
ByusingX∗logΦ=−∑
j ∂χj
∂xj ,wecanobtain:
α(X∗)= (DlogΦ)(X∗) ⇒
LX∗α=0 DX∗logΦ=−∑
j ∂χj
∂xj (136)
ByusingAX∗Y∗=−DY∗X∗,wecandevelop:
AX∗ (
∂
∂xj )
=−D ∂
∂xj X∗=−∑
i ∂χi
∂xj ∂
∂xi (137)
As f(X)=AX∗,o andq(X)=X∗o:
Tr(f(X))=Tr(AX∗,o)=−∑
i ∂χi
∂xi (o)=α(X∗0)=α0(q(X)) (138)
Ifweuse thatLX∗α=0∀X∈ g, thenweobtain:
(Dα)(X∗,Y∗)=(DY∗α)(X∗)=−(AY∗α)(X∗)=−AY∗ (α(X∗))+α(AY∗X∗)=α(AY∗X∗) (139)
79
Differential Geometrical Theory of Statistics
- Titel
- Differential Geometrical Theory of Statistics
- Autoren
- Frédéric Barbaresco
- Frank Nielsen
- Herausgeber
- MDPI
- Ort
- Basel
- Datum
- 2017
- Sprache
- englisch
- Lizenz
- CC BY-NC-ND 4.0
- ISBN
- 978-3-03842-425-3
- Abmessungen
- 17.0 x 24.4 cm
- Seiten
- 476
- Schlagwörter
- Entropy, Coding Theory, Maximum entropy, Information geometry, Computational Information Geometry, Hessian Geometry, Divergence Geometry, Information topology, Cohomology, Shape Space, Statistical physics, Thermodynamics
- Kategorien
- Naturwissenschaften Physik