The Joint Block Diagonalization (JBD) problem: a tensor framework. Dimitri Nion & Lieven De Lathauwer K.U. Leuven, Kortrijk campus, Belgium E-mails: Dimitri.Nion@kuleuven-kortrijk.be Lieven.DeLathauwer@kuleuven-kortrijk.be TDA 2010 , Monopoli, Italy, September 13-17, 2010 1
Joint-Block-Diagonalization (JBD): model I D 1 K 0 T X A L 1 L 1 L R 1 K D D RK 11 0 = I A A X T A 1 R 0 L R 1 D R 1 R T H X AD A ( N ) or X AD A ( N ), k 1 ,..., K k k k k k k JBD is a generalization of JD (Joint Diagonalization)/INDSCAL D K c 1 c R 0 0 X R K + … + T = = A a 1 a R 0 X 0 a R A I 1 a 1 D 1 2
JBD : ambiguities I D 0 1 K T L 1 L R A L 1 X 1 D D RK 0 K 11 Z -1 Z -T Z T Z = T I 0 A A A D L R X R 1 1 R 1 R ~ T ~ K = { D } A A k k 1 Observation: if you choose Z arbitrary, you lose the JBD structure. Question: what is the structure of Z such that the JBD model is still valid? 3
JBD : essential uniqueness Z K ΛΠ The JBD of X is said essentially unique if k k 1 Λ an arbitrary block-diagonal matrix, Π an arbitrary block-wise permutation matrix. 1 0 Λ 0 1 1 1 0 Λ ~ ~ ~ = A A A 0 1 2 A A A 1 2 3 1 1 2 3 0 Λ 0 3 1 Solving a JBD problem Estimation of {Span( A r )} r=1,…,R in an arbitrary order 4
JBD: State of the art JBD is becoming popular signal processing tool in applications such as: Blind Source Separation (BSS) of convolutive mixtures in the time-domain, Independent Subspace Analysis. Two approaches in the literature: Approach 1: Unitary-JBD [Abed Meraim and Belouchrani, 2004] « A is a square unitary matrix » ( A T A = I ) Approach 2: Non-Unitary JBD Approach 2.1: [H. Ghennioui, N. Thirion-Moreau, E. Moreau, 2008, 2010] « A is tall and full column-rank. » Indeed, their approach only works if A is a square non-unitary matrix. Approach 2.2: This talk « A can be a tall, square or fat non-unitary matrix » JBD is a particular instance of Block-Component-Decompositions Computation by a gradient-based algorithm In the square case, better performance than 2.1 5
Joint-Block-Diagonalization : state of the art (1) Approach 1: Unitary-JBD [Abed Meraim and Belouchrani, 2004] A is square unitary matrix ( A T A = I ) A D k A T (+ N k ) A T X k A N k A T ) X k = A D A = D k + (A N K K 2 2 T T max bdiag ( A X A ) or min offbdiag ( A X A ) k k F F A A k 1 k 1 X K = X 1 A T A 6
Joint-Block-Diagonalization : state of the art (2) Approach 2: Non-Unitary-JBD [H. Ghennioui, N. Thirion-Moreau, E. Moreau, 2008, 2010] A is tall and full column-rank (Let B = A , then BA = I ) BX k B T = D k + (B N k B T ) X k = A D A D k A T (+ N k ) BX K K 2 2 T T max bdiag ( BX B ) or min offbdiag ( BX B ) k k F F B B k 1 k 1 X K = X B T 1 B 7
Joint-Block-Diagonalization : state of the art (3) Approach 2: Non-Unitary-JBD [H. Ghennioui, N. Thirion-Moreau, E. Moreau, 2008, 2010] 2 gradient-descent based algorithms, JBD OG and JBD ORG to solve K 2 T min offbdiag ( BX B ) ( 1 ) off k F B k 1 Drawbacks of approach 2: B =0 is a trivial minimizer Under-determined case ( A fat, I<N) not handled, since it is assumed that BA = I Indeed, the over-determined case ( A tall, I>N) is not successfully handled either T T T because if is solution of an exact JBD problem, i.e., B [ B , B ] A 1 2 T T T ( BX B ) ( BAD A B ) ( D ) 0 , offbdiag offbdiag offbdiag k k k k ~ T T T then is also solution of (1) but not solution of the JBD problem B [( A ) , B ] 2 ~ 0 0 ~ T T T T because is not full rank. Idem for B [ B ( A ) , B ] B A 1 2 X X 8
Joint-Block-Diagonalization : our contributions Starting point: the gradient-descent based algorithms JBD OG and JBD ORG of [H. Ghennioui, N. Thirion-Moreau, E. Moreau, 2008, 2010] for Non-Unitary JBD can only handle the case where A is square (I=N). Main motivation: Propose a novel technique to solve non-unitary JBD problems that can handle exactly-, over- and under- determined cases (i.e., A may be square, tall or fat). Main contributions: Formulate JBD as a tensor decomposition fitting problem; Build a Conjugate Gradient (CG) algorithm to compute the tensor decomposition; In the over-determined case, build a good starting point for any JBD algorithm; Application: blind source separation via Independent Subspace Analysis (ISA). 9
JBD in tensor format : link to BCD D 0 1 K X L 1 L R T A K 1 = D D RK 0 11 I A A X 1 R T 1 A 0 R D R 1 X R L r T A = r I A D r r r 1 R X D A A r 1 r 2 r r 1 JD particular case of candecomp-parafac R JBD particular case of BCD-(L r , M r , .), (« Block- A X D * A (or Component-Decomposition in rank-(L r , M r , .) terms ») r 1 r 2 r r 1 in case of hermitian symmetry) 10
JBD in tensor format : conditions for essential uniqueness R BCD-(L r , M r , .) : where A =[ A 1 ,…, A R ] is I by N X D A B r 1 r 2 r B =[ B 1 ,…, B R ] is J by Q r 1 Theorem [De Lathauwer, 2008] : Suppose that rank( A )=N, rank( B )=Q, K>2 and that the tensors { D r } r=1,…,R are generic, then the BCD-(L r , M r , .) of X is essentially unique (Sufficient condition). R X D JBD : A =[ A 1 ,…, A R ] is I by N A A where r 1 r 2 r r 1 The same theorem can be invoked (the proof still holds with A instead of B ) In summary, it means that JBD is generically unique if K>2 and rank( A )=N This is only a sufficient condition: uniqueness still holds but is harder to prove in several cases where the condition is not satisfied. For instance, uniqueness may still hold when rank( A )=I ( A fat, I<N) 11
JBD in tensor format : cost function R 2 N N X D min A A with LS r 1 r 2 r F R D , A r r 1 1 r K R 2 T N N X A D A with k k k r kr r F 1 1 k r 2 . N with N X ( A A ) D F Where: . A A = [ A 1 A 1 , …, A R X X A R ] is the Khatri-Rao product (block-wise Kronecker product) X and N are the I 2 xK matrix unfoldings of the IxIxK tensors X and N , resp. 12
JBD in tensor format : derivation of the gradient Real-valued data K T T ( ) 2 ( N AD N AD ) , T X AD A N A LS k k k k k k k k 1 . T ( ) 2 ( A A ) N D LS K Complex-valued data, * T * * H ( ) ( N A D N A D ) , standard symmetry * LS k k k k A k 1 T . X AD A N H ( ) ( A A ) N k k k * LS D Complex-valued data, K hermitian symmetry H H ( ) ( N AD N AD ) , * LS k k k k A k 1 H X AD A N . * T k k k ( ) ( A A ) N * LS D 13
Conjugate Gradient algorithm for JBD: JBD-CG (stop criterion not satisfied) while 1 - Compute and from and D A A D 2 - Update search directions 2.1 - Compute (e.g. with stabilized Polak - Ribière formula) 2.2 - New search directions : d d A A A d d D D D 3 - Joint Exact Line Search : ( , ) arg min ( A d , D d ) A D LS A A D D , A D 4 - Update unknwons : A A d A A D D d D D End 14
Step 3 of JBD-CG : Joint Exact Line Search min ( A d , D d ) LS A A D D , A D . 2 ( A d , D d ) ( A d ) ( A d ) ( D d ) X A A D D A A A A D D LS F 2 Q ( ) 2 Q ( ) Q ( ) ( 1 ) D 2 A D 1 A 0 A LS ( , ) Q ( ) Solve A D 0 1 A ( 2 ) D Q ( ) D 2 A 2 Q ( ) Q ( ) Q ( ) Substitute (2) in (1): 1 A 0 A 2 A ( ) ( 3 ) A LS ( ) Q 2 A Minimize (3) w.r.t. (degree 7 polynomial rooting) A Given , is given by (2) A D 15
Recommend
More recommend