Seite - 418 - in Differential Geometrical Theory of Statistics
Bild der Seite - 418 -
Text der Seite - 418 -
Entropy2016,18, 425
(d) (e) (f)
(g) (h) (i)
(j) (k) (l)
Figure6. (aâl)With thesetupofFigures4and5,generatedfamiliesofcurvesbyvaryingthevertical
VâFM part of the initialmomentum Ο0 â TâFM but keeping the base point and frame u0 ïŹxed.
Theverticalpartallowsvaryingdegreeofâtwistingâof thecurve.
6.2. LDDMMLandmarkEquations
Weheregiveaexampleof theMPPequationsusingtheïŹnitedimensional landmarkmanifolds
thatarisefromrightinvariantmetricsonsubsetsofthediffeomorphismgroupintheLargeDeformation
DiffeomorphicMetricMapping(LDDMM)framework[8]. TheLDDMMmetriccanbeconveniently
expressedas a cometric, and, using a rank-deïŹcient innerproduct gFkM asdiscussed in Section 5,
wecanobtainareductionof thesystemofequations to2(2N+2Nk)comparedto2(2N+(2N)2)with
N landmarks inR2.
Let {p1, . . . ,pN} be landmarks in a subsetΩ â Rd. The diffeomorphism group Diff(Ω)
acts on the left on landmarkswith the action Ï.{p1, . . . ,pN} = {Ï(p1), . . . ,Ï(pN)}. In LDDMM,
aHilbert spacestructure is imposedona linearsubspaceVofL2(Ω,Rd)usingaself-adjointoperator
L :VâVââL2(Ω,Rd)anddeïŹningthe innerproduct ă·, ·ăV by
ăv,wăV= ăLv,wăL2 .
Under sufïŹcient conditions on L,V is reproducing and admits a kernelK inverse to L. K is
aGreenâskernelwhenL isadifferentialoperator,orK canbeaGaussiankernel. TheHilbert structure
onVgivesaRiemannianmetriconasubsetGVâDiff(Ω)bysettingâvâ2Ï=âvâŠÏâ1â2V; i.e., regarding
ă·, ·ăV aninnerproductonTIdGV andextendingthemetric toGV byright-invariance. ThisRiemannian
metricdescends toaRiemannianmetriconthe landmarkspace.
418
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