Statistics on Lie groups: using the pseudo-Riemannian framework? Nina Miolane, Xavier Pennec Asclepios Team, INRIA Sophia Antipolis Max Ent β 2014 Nina Miolane - Statistics on Lie groups
Models with Lie groups Articulated models Shape models Robotics Computational Medicine Reference ο¦ 1 π»π· π Spherical arm ο¦ 5 Patient 1 ο¦ 2 ο¦ 4 Patient 5 ο¦ 3 http://www.societyofrobots.com/ Patient 4 Patient 3 Patient 2 Paleontology Computational Anatomy π»π(π) 2006, Nature Publishing Group, Genetics ο Statistics 10/17/2014 2 Nina Miolane - Statistics on Lie groups
Models with Lie groups Articulated model of spine: Statistics on spines: Vertebra π, π’ β ππΉ(3) Mean, PCAβ¦ Rotation Matrix translation vector (π, π’) PCA : Modes 1 to 4 Computational Anatomy : model and analyze the variability of human anatomy 10/17/2014 3 Nina Miolane - Statistics on Lie groups
New statistics on Lie groups? On Lie group π»π(π) Consequence for model Vertebra : π, π’ β ππΉ(3) (space of transformations) π π , π π π π , π π ? ( πΊ π +πΊ π , π π +π π π π , π π πΊ π +πΊ π π π +π π π ) ( π ) , π π ? π π , π π π»π π Mean not on ππΉ(3) , thus not a Average pose of two vertebrae is rigid transformation not a vertebra Usual statistics: linear Need new statistics! Lie groups: not linear in general 10/17/2014 4 Nina Miolane - Statistics on Lie groups
Outline Can we use the pseudo-Riemannian framework to define statistics on Lie groups? 1. Pseudo-Riemannian structures on Lie groups 2. Algorithm to compute bi-invariant pseudo-metrics 3. Results on selected Lie groups 10/17/2014 5 Nina Miolane - Statistics on Lie groups
Outline Can we use the pseudo-Riemannian framework to define statistics on Lie groups? 1. Pseudo-Riemannian structures on Lie groups 2. An algorithm to compute bi-invariant pseudo-metrics 3. Results on selected Lie groups 10/17/2014 6 Nina Miolane - Statistics on Lie groups
(Pseudo-) Riemannian structure on Lie groups Manifold π΅ Differential structure + algebraic + metrical Lie group (Pseudo-) Riemannian manifold (π―,β) (π΅, <, >) Metric <, > : collection of positive definite inner products Pseudo-metric <, > : collection of definite inner products Consistency of the structures requires bi-invariant pseudo-metric Quadratic Lie group (π―,β, <, >) 10/17/2014 7 Nina Miolane - Statistics on Lie groups
Statistics on Lie groups : the mean π π π π ο Requirements for mean of data set Consistency Computability Riemannian structure π mean of π π π ο Bi-invariant metric Conditions on π π π FrΓ©chet mean = group mean π β such that exists π FrΓ©chet mean: definition with and is computability conditions unique? π β π mean of β β π π π Group exponential barycenter [1] =group mean [1] Pennec & Arsigny. Exponential barycenter of the canonical Cartan connection and invariant means on Lie groups. (2012) 10/17/2014 8 Nina Miolane - Statistics on Lie groups
Bi-invariant metric? On all Lie groups? NO! Existence of a bi-invariant metric ? On Lie groups that have a group mean? NO! Characterization of Lie groups with π»π(π) : bi-invariant metric by Cartan [2] : β’ unique (a.e.) group mean compact & abelian β’ no bi-invariant metric ο Group mean not characterized by a bi-invariant metric [1] [1] Pennec & Arsigny. Exponential barycenter of the canonical Cartan connection and invariant means on Lie groups. (2012) [2] Elie Cartan. La thΓ©orie des groups finis et continus et lβanalyse situs. (1952) 10/17/2014 9 Nina Miolane - Statistics on Lie groups
Statistics on Lie groups : the mean π π π π ο Requirements for mean of data set Consistency Computability Riemannian structure π mean of π π π FrΓ©chet mean: definition with Conditions computability conditions on π π π π β such that exists π Bi-invariant metric and is ο FrΓ©chet mean = group mean unique? π β π mean of β β π π π Generalize Riemannian to pseudo-Riemannian Pseudo-Riemannian structure Group exponential barycenter [1] =group mean Bi-invariant pseudo-metric [1] Pennec & Arsigny. Exponential barycenter of the canonical Cartan connection and invariant means on Lie groups. (2012) 10/17/2014 10 Nina Miolane - Statistics on Lie groups
Bi-invariant pseudo-metrics? On all Lie groups? NO! Existence of a bi-invariant pseudo-metric ? On Lie groups that have a group mean? ππΉ(π) : Characterization of Lie groups with β’ unique (a.e.) group mean bi-invariant pseudo-metric β’ bi-invariant pseudo-metric ? by Medina & Revoy [3] Group mean characterized by a bi-invariant pseudo-metric ? [3] Medina & Revoy. Groupes de Lie munis de pseudo-metriques bi-invariantes. (1982) 10/17/2014 11 Nina Miolane - Statistics on Lie groups
Outline Can we use the pseudo-Riemannian framework to define statistics on Lie groups? 1. Pseudo-Riemannian structures on Lie groups 2. An algorithm to compute bi-invariant pseudo-metrics 3. Results on selected Lie groups 10/17/2014 12 Nina Miolane - Statistics on Lie groups
From Lie group to Lie algebra Compute bi-invariant pseudo-metrics on finite dimensional connected Lie group π― π¬π΄ π . π bi-invariant : < π, π > π =< π¬π΄ π . π, π¬π΄ π . π > π΄ π π =< πΈπ β . π£, πΈπ β . π€ > π β π π΄ π where π β π = β β π and π β π = π β β π π π΄ π π = π β π π π¬π΄ π . π Compute bi-invariant pseudo-metrics on its Lie algebra π = π π π, +, . , , π bi-invariant: < π¦, π§ π₯ , π¨ > π +< π§, π¦, π¨ π₯ > π = 0 Structure of quadratic π ? 10/17/2014 13 Nina Miolane - Statistics on Lie groups
Lie algebra representations Structure of quadratic π ? Study adjoint representation of π₯ Representation π½ of π on πΎ : Lie algebra homomorphism π: π₯ β¦ π₯πͺ(π) Ex.: Homogeneous representation of ππΉ 3 on β 4 : π: ππΉ 3 β¦ π₯πͺ β 4 1 . π¦ s.t. π π’ 1 = π. π¦ + π’ π, π’ β¦ π π’ 0 1 1 0 Subrepresentation: subspace of π stable by π π₯ Subrepresentation decomposition: π = πΆ 1 β π₯ β¦ β π₯ πΆ π with πΆ π subrepresentations Indecomposable subrepresentation: not a sum of subrepresentations Adjoint representation ππ of π (on itself: πΎ = π ) ππ: π₯ β¦ π₯πͺ(π₯) π¦ β¦ ππ π¦ = π¦,β π₯ π₯ β π₯ β¦ β π₯ πΆ π π₯ : decomposition into indecomposable subrepresentations π₯ = πΆ 1 10/17/2014 14 Nina Miolane - Statistics on Lie groups
Structure of quadratic π₯ π₯ π₯ π₯ π₯ πΆ 1 πΆ π πΆ π π = π β π» β π = π β π β π» β π =1-dim. π =simple πͺ π πͺ π πͺ π πͺ π Th. (Medina & Revoy): Structure of quadratic π π₯ β π₯ β¦ β π₯ B N π₯ has indecomposable B π π₯ s.t.: Adjoint representation decomposition π₯ = B 1 π simple or 1-dim. Type (1): πͺ π π = π β π β π β double extension of π quadratic by π of Type (1) Type (2): πͺ π 10/17/2014 15 Nina Miolane - Statistics on Lie groups
Bi-invariant pseudo-metric on quadratic π₯ β² + β― + π π > π₯ π₯ < π 1 + β― + π π , π 1 β² > πΆ 1 + β― +< π π , π π β² > πΆ π =< π 1 , π 1 π₯ π₯ π₯ πΆ 1 πΆ π πΆ π π = π β π» β π = π β π β π» β π =1-dim. π =simple πͺ π πͺ π πͺ π πͺ π π = π³ππππππ(π, π β² ) < π, πβ² > πͺ π < π + π + π, π β² + π β² + π β² > πͺ π π = < π, π β² > πΏ + π π β² + πβ²(π) < π, π β² > πͺ π π = π. πβ² < π + π, π β² + π β² > πͺ π π = π π β² + πβ²(π) 10/17/2014 16 Nina Miolane - Statistics on Lie groups
Algorithm π₯ π₯ π₯ π₯ πΆ 1 πΆ π πΆ π π = π β π» β ? π = π β π β π» β ? Else : EXIT π =1-dim. ? π =simple? πͺ π πͺ π πͺ π πͺ π If algorithm finishes: expression of a bi-invariant pseudo-metric on π₯ If EXIT: no bi-invariant pseudo-metric on π₯ 10/17/2014 17 Nina Miolane - Statistics on Lie groups
Outline Can we use the pseudo-Riemannian framework to define statistics on Lie groups? 1. Pseudo-Riemannian structures on Lie groups 2. An algorithm to compute bi-invariant pseudo-metrics 3. Results on selected Lie groups 10/17/2014 18 Nina Miolane - Statistics on Lie groups
Recommend
More recommend