Seite - 123 - in Differential Geometrical Theory of Statistics
Bild der Seite - 123 -
Text der Seite - 123 -
Entropy2016,18, 254
For instance, ifG is thegroupofEuclideantransformations,werecover theclassicalEuclidean
tensors. Hence, each G-tensor can be identified with an orbit of G within the space of the
tensorcomponents.
3.MomentumasAffineTensor
LetM be a differential manifold of dimension n and G a Lie subgroup ofAf f(n). In the
applications tophysics,Mwillbe forus typically thespace-timeandGasubgroupofAf f(n)witha
physicalmeaning intheframeworkofclassicalmechanics (Galileo’sgroup)orrelativity (Poincaré’s
group). Thepointsof thespace-timeMareeventsofwhichthecoordinateX0 is thetime tandXi= xi
for i runningfrom1to3gives theposition.
The tangent space toM at thepointX equippedwitha structureof affine space is called the
affine tangent spaceand isdenotedATXM. Its elementsare called tangentpoints atX. The setof
affineformsontheaffinetangentspace isdenotedA∗TXM.Wecallmomentumabilinearmapμ:
μ :TXM×A∗TXM→R : (−→V ,Ψ) →μ(−→V ,Ψ)
It isamixed1-covariantand1-contravariantaffine tensor. Taking intoaccount thebilinearity, it is
represented inanaffineframe f by:
μ( −→
V ,Ψ)=(χKβ+ΦαLαβ)V β
whereKβ andLαβ are thecomponentsofμ in theaffineframe f or,equivalently, thecoupleμ=(K,L)
of therowK collecting theKβ andthen×nmatrixLofelementsLαβ. Owingto (2), the transformation
lawisgivenbythe inducedactionofAf f(n):
K′=KP−1, L′=(PL+CK)P−1 (3)
If theaction is restrictedto thesubgroupG, themomentumμ isaG-tensor.
Ontheotherhand,havea lookto theLiealgebragofG, that is thesetof infinitesimalgenerators
Z= da=(dC,dP)with a∈G. Letus identify thespaceof themomentumcomponentsμ=(K,L) to
thedualg∗of theLiealgebra thanks to thedualpairing:
μZ=μda=(K,L)(dC,dP)=KdC+Tr(LdP) (4)
Weknowthat thegroupactson itsLiealgebrabytheadjoint representation:
Ad(a) :g→g :Z′ →Z=Ad(a)Z′= aZ′a−1 .
AsG isagroupofaffinetransformations,any infinitesimalgeneratorZ is representedby:
Z˜= dP˜= d (
1 0
C P )
= (
0 0
dC dP )
.
Then Z˜= P˜ Z˜′ P˜−1 leads to:
dC=P(dC′−dP′P−1C), dP=PdP′P−1 . (5)
Thisadjoint representation induces thecoadjoint representationofG ing∗definedby:
(Ad∗(a)μ′)Z=μ′(Ad(a−1)Z) .
123
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