Page - 105 - in Differential Geometrical Theory of Statistics
Image of the Page - 105 -
Text of the Page - 105 -
Entropy2016,18, 386
Butasλ(θ)>0,wecanconsider 1
λ(θ) as the secondderivativeof a functionΦ(θ) suchthat:
∂logpθ(x)
∂θ = ∂2Φ(θ)
∂θ2 [h(x)−θ] (A16)
Fromwhichwededuce that:
(x)= logpθ(x)− ∂Φ(θ)∂θ [h(x)−θ]−Φ(θ) (A17)
Isan independentquantityofθ.Adistinguished functionwillbe thengivenby:
pθ(x)= e ∂Φ(θ)
∂θ [h(x)−θ]+Φ(θ)+ (x) (A18)
With thenormalizingconstraint +∞
−∞ pθ(x)dx=1.
These twoconditionsaresufficient. Indeed, reciprocally, let three functionsΦ(θ),h(x)and (x)
thatwehave, forany
θ : +∞
−∞ e ∂Φ(θ)
∂θ [h(x)−θ]+Φ(θ)+ (x)dx=1 (A19)
Thenthe function isdistinguished:
θ+ ∂logpθ(x)
∂θ
+∞
−∞ [
∂pθ(x)
∂θ ]2 dx
pθ(x) = θ+λ(x) ∂2Φ(θ)
∂θ2 [h(x)−θ] (A20)
Ifλ(x) ∂2Φ(θ)
∂θ2 =1, when 1
λ(x) = +∞
−∞ [
∂logpθ(x)
∂θ ]2
pθ(x)dx=(σA) 2 (A21)
Thefunction is reducedtoh(x)andthen isnotdependentofθ.
Wehavethenthe followingrelation:
1
λ(x) = +∞
−∞ ( ∂2Φ(θ)
∂θ2 )2
[h(x)−θ]2e∂Φ(θ)∂θ (h(x)−θ)+Φ(θ)+ (x)dx (A22)
Therelation isvalid foranyθ,wecanderiveprefiousequationwithrespectwithθ:
+∞
−∞ e ∂Φ(θ)
∂θ (h(x)−θ)+Φ(θ)+ (x) ( ∂2Φ(θ)
∂θ2 )
[h(x)−θ]dx=0 (A23)
Wecandivideby ∂2Φ(θ)
∂θ2 because itdoesnotdependonx.
Ifwederiveagainwithrespect toθ,wewillhave:
+∞
−∞ e ∂Φ(θ)
∂θ (h(x)−θ)+Φ(θ)+ (x) ( ∂2Φ(θ)
∂θ2 )
[h(x)−θ]2dx= +∞
−∞ e ∂Φ(θ)
∂θ (h(x)−θ)+Φ(θ)+ (x)dx=1 (A24)
105
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