100 likes | 125 Views
Learn the unique decomposition of matrices using SVD and its applications in determining matrix singularity, computing inverses, and matrix conditions. Gain insights into eigenvalues and eigenvectors through illustrative examples.
E N D
Singular Value Decomposition (SVD)(see Appendix A.6, Trucco & Verri) CS485/685 Computer Vision Prof. George Bebis
SVD • Any real m x nmatrix A can be decomposed uniquely: • U is m x nand column orthonormal (UTU=I) • D is n x nanddiagonal • σi are called singularvalues of A • It is assumed that σ1 ≥ σ2 ≥ … ≥ σn ≥ 0 • V is n x nand orthonormal (VVT=VTV=I)
SVD (cont’d) • If m=n, then: • U is n x nand orthonormal (UTU=UUT=I) • D is n x nanddiagonal • V is n x nand orthonormal (VVT=VTV=I)
SVD (cont’d) • The columns of U are eigenvectors of AAT • The columns of V are eigenvectors of ATA • If λi is an eigenvalue of ATA (or AAT), then λi =σi2 for square matrices: A=PΛP-1
SVD - Example U = (u1u2 . . . un) V = (v1v2 . . . vn) D
λ1 λ2 λ3 SVD – Another Example The eigenvalues of AAT, ATA are: The eigenvectors of AAT, ATA are:
SVD properties • A square (n × n) matrix A is singular iff at least one of its singular values σ1, …, σn is zero. • The rank of matrix A is equal to the number of nonzerosingular values σi
Matrix “condition” • SVD gives a way of determining how singular A is. • The conditionof Ameasures the degree of singularity of A: cond(A)= • (ratio of largest singular value to its smallest singular value) • Matrices with large condition number are called • ill conditioned.
Computing A-1 using SVD • If A is a n x n nonsingular matrix, then its inverse can be computed as follows: easy to compute! (UTU=UUT=I or UT=U-1 and VTV=VVT=I or VT=V-1)
Computing A-1 using SVD (cont’d) • If A is singular (or ill-conditioned), we can use SVD to approximate its inverse as follows: ? where (t is a small threshold)