Page - 201 - in Differential Geometrical Theory of Statistics
Image of the Page - 201 -
Text of the Page - 201 -
Entropy2016,18, 433
Thefirst critiquearises fromrequirement (5).
Fromtheviewpointoffiberbundlestherequirement(5) isuseless.ConsidertheCartesianproduct
E=Θ×Ξ.
That is thesamethingas the trivialfiberbundle
E (θ,ξ)→π(θ,ξ)= θ∈Θ.
ThereforePθ is therestrictionto thefiberEθ of the functionP.
TheSecondCritique
Thesecondcritiqueemerges fromtherequirement (1).
This requirement (1) is toorestrictive. It excludesmanyinterestingcompactmanifoldssuchas
flat tori, euclideansphere, compactLiegroups.
TheThirdCritique
From the viewpoint of the differential topology the requirement (6) may be damage to the
topology ofΘ. When the Fisher information g is singular its kernel is in involution. Thus the
topological-geometrical informationthatarecontainedingare transverse to thedistributionKer(g).
IfKer(g) is completely integrable thentopologicalandgeometrical informationswhicharecontained
ingare transversal to the foliationKer(g). SeePartAof thispaper. Thisends thecriticisms.
Tomotivate for deleting the requirement (1)we construct a compact statisticalmodelwhich
satisfiesallof therequirementsexcept therequirement (1).
LetE bethe tangentbundleof thecircleS1.E is the trivial linebundle
S1×R (θ,t)→ θ∈S1.
Weconsider the fonctions f,FandPdefinedby
f(θ,t)= [sin2( t2θ
1+ t2 )cos2( θ
4 )e−t
2
+ π
e2 t2],
F(θ)= ∫ +∞
−∞ e−f(θ,t)dt,
P(θ,t)= e−f(θ,t)
F(θ) .
The functionP(θ,t)has the followingproperties
(i) (i) :P(θ,t) is smooth,
(ii) P(0,t)=P(2π,t) ∀t∈R,
(iii) the ddθ commuteswith ∫
R ,
(iv) P(θ,t)≤1 ∀(θ,t)∈S1×R,
(v) if0< θ,θ∗<2π thenPθ=P∗θ if andonly ifθ= θ ∗,
(vi) ∫+∞
−∞ P(θ,t)dt=1.
Thesepropertiesshowthat there isaonetoonecorrespondencebetweenthecircleS1 andasubset
of probabilitydensities inR. ThusS1 is a compact 1-dimensionalmanifoldof probabilities in the
measurableset (R,β(R)).Hereβ(R) is the familyofBorel subsetsofR.
So (S1,P) isacompactparametricmodel for (R,β(R)).
201
Differential Geometrical Theory of Statistics
- Title
- Differential Geometrical Theory of Statistics
- Authors
- Frédéric 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