orthogonal similarity reduction of any symmetric matrix
play

Orthogonal similarity reduction of any symmetric matrix into a - PowerPoint PPT Presentation

Mar 05, 2023 •462 likes •1.53k views

Orthogonal similarity reduction of any symmetric matrix into a diagonal-plus-semiseparable one with free choice of the diagonal Ellen Van Camp, Raf Vandebril, Marc Van Barel and Nicola Mastronardi I. Algorithms Orthogonal similarity

  1. × × × 0 ⊠ ⊠ ⊠ × × × 0 ⊠ ⊠ ⊠ × × × 0 ⊠ ⊠ ⊠ 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ × ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 0 0 0 0 ⊠ ⊠

  2. × × × 0 ⊠ ⊠ ⊠ × × × 0 ⊠ ⊠ ⊠ × × × 0 ⊠ ⊠ ⊠ 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 0 0 0 0 ⊠ ⊠

  3. × × × 0 ⊠ ⊠ ⊠ × × × 0 ⊠ ⊠ ⊠ × × × 0 ⊠ ⊠ ⊠ 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊗ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 0 0 0 0 ⊠ ⊠

  4. × × × 0 ⊠ ⊠ ⊠ × × × 0 ⊠ ⊠ ⊠ × × × 0 ⊠ ⊠ ⊠ 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ × ⊠ ⊠ ⊠ ⊠ ⊠ ⊠

  5. × × × ⊠ ⊠ ⊠ ⊠ × × × ⊠ ⊠ ⊠ ⊠ × × × ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ × × ⊠ ⊠ ⊠ ⊠ ⊠

  6. × × × ⊠ ⊠ ⊠ ⊠ × × × ⊠ ⊠ ⊠ ⊠ × × × ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ × ⊠ ⊠ ⊠ ⊠ ⊠ ⊠

  7. × × × ⊠ ⊠ ⊠ ⊠ × × × ⊠ ⊠ ⊠ ⊠ × × × ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠

  8. 3. Reduction to diagonal-plus-semiseparable form with free choice of the diagonal Any symmetric matrix can be transformed into a diagonal-plus-semiseparable one where the diagonal can be chosen in advance, by means of orthogonal similarity transformations in order O ( 4 3 n 3 ) . Definition The sum of a symmetric semiseparable matrix and a diagonal matrix is called a diagonal-plus-semiseparable matrix. So choose a diagonal d = [ d 1 , d 2 , . . . , d n ] .

  9. A first algorithm × × × × × × × × × × × × × × × × × × × × × − → D + S × × × × × × × × × × × × × × × × × × × × × × × × × × × ×

  10. Step 1 × × × × × × × 0 0 0 0 0 0 0 × × × × × × × 0 0 0 0 0 0 0 × × × × × × × 0 0 0 0 0 0 0 + × × × × × × × 0 0 0 0 0 0 0 × × × × × × × 0 0 0 0 0 0 0 × × × × × × × 0 0 0 0 0 0 0 × × × × × × × 0 0 0 0 0 0 d 1

  11. × × × × × × ⊗ 0 0 0 0 0 0 0 × × × × × × ⊗ 0 0 0 0 0 0 0 × × × × × × ⊗ 0 0 0 0 0 0 0 + × × × × × × ⊗ 0 0 0 0 0 0 0 × × × × × × ⊗ 0 0 0 0 0 0 0 × × × × × × × 0 0 0 0 0 0 0 ⊗ ⊗ ⊗ ⊗ ⊗ × × 0 0 0 0 0 0 d 1

  12. × × × × × × 0 0 0 0 0 0 0 0 × × × × × × 0 0 0 0 0 0 0 0 × × × × × × 0 0 0 0 0 0 0 0 + × × × × × × 0 0 0 0 0 0 0 0 × × × × × × 0 0 0 0 0 0 0 0 × × × × × × × 0 0 0 0 0 0 0 0 0 0 0 0 × × 0 0 0 0 0 0 d 1

  13. × × × × × × 0 0 0 0 0 0 0 0 × × × × × × 0 0 0 0 0 0 0 0 × × × × × × 0 0 0 0 0 0 0 0 + × × × × × × 0 0 0 0 0 0 0 0 × × × × × × 0 0 0 0 0 0 0 0 × × × × × × ⊗ 0 0 0 0 0 0 0 0 0 0 0 0 ⊗ × 0 0 0 0 0 0 d 1

  14. Problem � � � � � � � � s 2 d 1 0 0 − s c s c csd 1 = c 2 d 1 − s 0 c d 1 s c csd 1 ⇒ ???

  15. Solution � � � � � � 0 − s c s d 1 c − s 0 c d 1 s c � � � � � � 1 0 − s c s c = d 1 − s 0 1 c s c � � 0 d 1 = 0 d 1

  16. × × × × × × 0 0 0 0 0 0 0 0 × × × × × × 0 0 0 0 0 0 0 0 × × × × × × 0 0 0 0 0 0 0 0 + × × × × × × 0 0 0 0 0 0 0 0 × × × × × × 0 0 0 0 0 0 0 0 × × × × × × ⊗ 0 0 0 0 0 0 0 0 0 0 0 0 ⊗ × 0 0 0 0 0 0 d 1

  17. × × × × × × 0 0 0 0 0 0 0 0 × × × × × × 0 0 0 0 0 0 0 0 × × × × × × 0 0 0 0 0 0 0 0 + × × × × × × 0 0 0 0 0 0 0 0 × × × × × × 0 0 0 0 0 0 0 0 × × × × × + ⊗ 0 0 0 0 0 0 d 1 0 0 0 0 0 ⊗ × 0 0 0 0 0 0 d 1

  18. × × × × × 0 0 0 0 0 0 0 ⊠ ⊠ × × × × × 0 0 0 0 0 0 0 ⊠ ⊠ × × × × × 0 0 0 0 0 0 0 ⊠ ⊠ + × × × × × 0 0 0 0 0 0 0 ⊠ ⊠ × × × × × 0 0 0 0 0 0 0 ⊠ ⊠ 0 0 0 0 0 0 d 1 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 0 0 0 0 0 d 1 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠

  19. × × × × × 0 0 0 0 0 0 0 ⊠ ⊠ × × × × × 0 0 0 0 0 0 0 ⊠ ⊠ × × × × × 0 0 0 0 0 0 0 ⊠ ⊠ + × × × × × 0 0 0 0 0 0 0 ⊠ ⊠ × × × × × 0 0 0 0 0 0 0 ⊠ ⊠ 0 0 0 0 0 0 d 1 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 0 0 0 0 0 d 2 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊞

  20. × × × × × 0 0 0 0 0 0 0 ⊠ ⊠ × × × × × 0 0 0 0 0 0 0 ⊠ ⊠ × × × × × 0 0 0 0 0 0 0 ⊠ ⊠ + × × × × × 0 0 0 0 0 0 0 ⊠ ⊠ × × × × × 0 0 0 0 0 0 0 ⊠ ⊠ 0 0 0 0 0 0 d 1 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 0 0 0 0 0 d 2 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠

  21. Step 3 × × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ × × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ × × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ + × × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ 0 0 0 0 0 0 d 1 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 0 0 0 0 0 d 2 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 0 0 0 0 0 d 3 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠

  22. × × × × ⊗ ⊗ ⊗ 0 0 0 0 0 0 0 × × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ × × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ + × × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊗ 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 1 ⊗ 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 2 ⊗ 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 3

  23. × × × × 0 0 0 0 0 0 0 0 0 0 × × × × ⊗ ⊗ ⊗ 0 0 0 0 0 0 0 × × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ + × × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ 0 ⊗ 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ d 1 0 ⊗ 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ d 2 0 ⊗ 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ d 3

  24. × × × × 0 0 0 0 0 0 0 0 0 0 × × × × 0 0 0 0 0 0 0 0 0 0 × × × × ⊗ ⊗ ⊗ 0 0 0 0 0 0 0 + × × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ 0 0 ⊗ 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ d 1 0 0 ⊗ 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ d 2 0 0 ⊗ 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ d 3

  25. × × × × 0 0 0 0 0 0 0 0 0 0 × × × × 0 0 0 0 0 0 0 0 0 0 × × × × 0 0 0 0 0 0 0 0 0 0 + × × × × ⊗ ⊗ ⊗ 0 0 0 0 0 0 0 0 0 0 ⊗ 0 0 0 0 0 0 ⊠ ⊠ ⊠ d 1 0 0 0 ⊗ 0 0 0 0 0 0 ⊠ ⊠ ⊠ d 2 0 0 0 ⊗ 0 0 0 0 0 0 ⊠ ⊠ ⊠ d 3

  26. × × × × 0 0 0 0 0 0 0 0 0 0 × × × × 0 0 0 0 0 0 0 0 0 0 × × × × 0 0 0 0 0 0 0 0 0 0 + × × × + ⊗ ⊗ ⊗ 0 0 0 0 0 0 d 1 0 0 0 ⊗ 0 0 0 0 0 0 ⊠ ⊠ ⊠ d 1 0 0 0 ⊗ 0 0 0 0 0 0 ⊠ ⊠ ⊠ d 2 0 0 0 ⊗ 0 0 0 0 0 0 ⊠ ⊠ ⊠ d 3

  27. × × × 0 0 0 0 0 0 0 0 0 ⊠ ⊠ × × × 0 0 0 0 0 0 0 0 0 ⊠ ⊠ × × × 0 0 0 0 0 0 0 0 0 ⊠ ⊠ + 0 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ d 1 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 1 0 0 0 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ d 2 0 0 0 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ d 3

  28. × × × 0 0 0 0 0 0 0 0 0 ⊠ ⊠ × × × 0 0 0 0 0 0 0 0 0 ⊠ ⊠ × × × 0 0 0 0 0 0 0 0 0 ⊠ ⊠ + 0 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ d 1 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊞ ⊠ ⊠ d 2 0 0 0 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ d 2 0 0 0 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ d 3

  29. × × × 0 0 0 0 0 0 0 0 0 ⊠ ⊠ × × × 0 0 0 0 0 0 0 0 0 ⊠ ⊠ × × × 0 0 0 0 0 0 0 0 0 ⊠ ⊠ + 0 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ d 1 ⊗ ⊗ 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ d 2 0 0 0 0 ⊗ 0 0 0 0 0 0 ⊠ ⊠ d 2 0 0 0 0 ⊗ 0 0 0 0 0 0 ⊠ ⊠ d 3

  30. × × × 0 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ × × × 0 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ × × × 0 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ + 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 1 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 2 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 2 0 0 0 0 0 0 0 0 0 0 0 ⊠ ⊠ d 3

  31. × × × 0 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ × × × 0 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ × × × 0 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ + 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 1 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 2 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊞ ⊠ d 3 0 0 0 0 0 0 0 0 0 0 0 ⊠ ⊠ d 3

  32. × × × 0 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ × × × 0 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ × × × 0 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ + 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 1 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 2 ⊗ 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 3 0 0 0 0 0 ⊗ 0 0 0 0 0 0 ⊠ d 3

  33. × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ + 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 1 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 2 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 3 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 3

  34. × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ + 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 1 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 2 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 3 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊞ d 4

  35. × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ × × × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ + 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 1 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 2 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 3 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 4

  36. A second algorithm Before the last step of the first algorithm: × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 0 0 0 0 0 d 1 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 0 0 0 0 0 d 2 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ + 0 0 0 0 0 0 d 3 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 0 0 0 0 0 d 4 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 0 0 0 0 0 d 5 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 6

  37. When applying the first algorithm starting with D = [ d 2 , d 3 , . . . , d n , ⋆ ] with ⋆ an arbitrary element, instead of [ d 1 , d 2 , . . . , d n ] , we get the following situation before the last step: × 0 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 0 0 0 0 0 d 2 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 0 0 0 0 0 d 3 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ + 0 0 0 0 0 0 d 4 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 5 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 6 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 7

  38. No last step necessary: 0 0 0 0 0 0 ⊞ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 1 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 2 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 3 + 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 4 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 5 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 6 0 0 0 0 0 0 ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ ⊠ d 7

  39. Any arbitrary symmetric matrix can be transformed into a symmetric diagonal-plus-semiseparable one with free choice of the diagonal by means of an orthogonal similarity transformation Q such that Qe 1 = e 1 .

  40. II. Accuracy −13 10 −14 10 −15 10 Tridiagonal Semiseparable Diagonal−plus−semiseparable −16 10 0 500 1000 1500 2000 2500

  41. III. Computational complexity −7 2.5 x 10 Tridiagonal Semiseparable Diagonal−plus−semiseparable 2 1.5 1 0.5 500 1000 1500 2000 2500

  42. IV. Convergence behavior of reduction algorithm into diagonal-plus-semiseparable form

  43. For the reduction to semiseparable form Eigenvalues are equidistant 1 : 200. 200 180 160 140 120 100 80 60 40 20 0 0 20 40 60 80 100 120 140 160 180 200

  44. Eigenvalues 1 : 100 and 1000 : 1100. 1200 1000 800 600 400 200 0 0 20 40 60 80 100 120 140 160 180 200

  45. For the reduction to diagonal-plus-semiseparable form Some notation A (0) = A A ( m ) m A ( m − 1) Q m Q T =   R T  A m 1 =  ( D + S ) m R 1 Q T = 1: m AQ 1: m where ( D + S ) m is a square diagonal-plus- semiseparable matrix of dimensions ( m + 1 ) × ( m + 1 ) .

  46. Lemma Q 1 : m < e n > = ( A − d m I )( A − d m − 1 I ) . . . ( A − d 1 I ) < e n >, for m = 1 , 2 , . . . and Q 1 : 0 = I.

  47. Proof. . For m = 0 : Q 1 : 0 < e n > = < e n > . . Suppose the theorem is true for m − 1, i.e., Q 1 : m − 1 < e n > = ( A − d m − 1 I ) . . . ( A − d 2 I )( A − d 1 I ) < e n > . m A ( m − 1 ) is of the form: The structure of Q T     × . . . × 0 . . . 0 0     . . . . ... . . . .     . . . .             × × 0 0 0 . . . . . .     Q T +     m     × × × 0 . . . . . . d 1         . . . . ... ...     . . . .     . . . .         × × × × . . . . . . d m Q T = H + m D

  48. Hence, m ( A ( m − 1) ) Q T H + Q T = m D Q T m ( Q T H + Q T 1: m − 1 AQ 1: m − 1 ) = m D ⇒ AQ 1: m − 1 − Q 1: m − 1 D = Q 1: m H Applying the former equality on < e n > and using the induction hypothesis, we derive that: ( AQ 1: m − 1 − Q 1: m − 1 D ) < e n > = Q 1: m H < e n > ( AQ 1: m − 1 − Q 1: m − 1 d m I ) < e n > = Q 1: m < e n > ⇒ ( A − d m I )( A − d m − 1 I ) . . . ( A − d 1 I ) < e n > = Q 1: m < e n >

  49. Lanczos-Ritz convergence behavior a) Lanczos-Ritz values Because AQ 1: m = Q 1: m A ( m ) equals:   R T  A m � ← Q 1: m |− − → � ← Q 1: m |− − → � � 1  . = A Q 1: m Q 1: m ( D + S ) m R 1 Hence, the eigenvalues of ( D + S ) m are the Ritz-values of A with respect to the subspace spanned by the columns of − → Q 1 : m .

  50. b) Connection with the Krylov subspace Some notation K m = < e n , Ae n , A 2 e n , . . . , A m e n >   ˜ hq T H h    0 × × × ˜ . . . H ∈ R ( n − m − 1) × ( n − m − 2)          0 × h ∈ R ( n − m − 1) × 1 Q m + 1 = .     . . .  ...    q ∈ R ( m +1) × 1 . . .    . . .     0 0 0 × ×

  51. We want to prove by induction that: col ( − → � � span Q 1 : m + 1 ) = K m + 1 . We have:   hq T h   × × × . . .     − → [ ← Q 1: m |− − →   0 × = Q 1: m ] Q 1: m +1 .     . .   ... . .   . .     0 0 × ×   × × × . . .   0 ×   ← Q 1: m h [1 , q T ] + − − →   = Q 1: m . . .   ... . .   . .     0 0 × ×

  52. Because: col ( − → � � = K m = < e n , Ae n , . . . , A m e n > • Q 1: m ) span − → • Q 1: m +1 < e n > = ( A − d m +1 I ) . . . ( A − d 1 I ) < e n > ← − ⇒ Q 1: m h ∈ K m +1 \K m We get: col ( − → � � = K m + 1 = < e n , Ae n , . . . , A m + 1 e n > span Q 1 : m + 1 )

  53. Theorem The eigenvalues of ( D + S ) m , the lower diagonal blocks that appear during the reduction algorithm, are the Lanczos-Ritz values of A with respect to the Krylov subspace K m .

Download Presentation
Download Policy: The content available on the website is offered to you 'AS IS' for your personal information and use only. It cannot be commercialized, licensed, or distributed on other websites without prior consent from the author. To download a presentation, simply click this link. If you encounter any difficulties during the download process, it's possible that the publisher has removed the file from their server.

Recommend

It V T V = L In mm LV = I with L = V T LEFT INVERSE COMPSCI 527 Computer Vision The

It V T V = L In mm LV = I with L = V T LEFT INVERSE COMPSCI 527 Computer Vision The

Orthogonal Matrices The Left Inverse of an Orthogonal Matrix M E m q i = v T i p (Finding coefficients q i is easy!) O_O Rewrite v T i v j = ij in matrix form V = [ v 1 , . . . , v n ] 2 R m n It V T V = L In mm LV = I with L =

467 views • 9 slides
Matrix Orthogonal Polynomials and Time and Band limiting Mirta M. Castro Smirnova Universidad de

Matrix Orthogonal Polynomials and Time and Band limiting Mirta M. Castro Smirnova Universidad de

Introduction Matrix Orthogonal Polynomials and band and time limiting Matrix Orthogonal Polynomials and Time and Band limiting Mirta M. Castro Smirnova Universidad de Sevilla, Espa na Joint work with F. Alberto Gr unbaum, University of

1.07k views • 79 slides
= an orthogonal matrix T X X l l Cov( , ) i k ij kj j = 1 = X F l Cov( , )

= an orthogonal matrix T X X l l Cov( , ) i k ij kj j = 1 = X F l Cov( , )

Factor analysis (cf. section 9.3) In matrix notation the factor model takes the form: An observable random vector X of dimension p has mean X LF = + vector and covariance matrix L Here is the p x m matrix

98 views • 6 slides
Principal Component Analysis and Autoencoders Shuiwang Ji Department of Computer Science &amp;

Principal Component Analysis and Autoencoders Shuiwang Ji Department of Computer Science &amp;

Principal Component Analysis and Autoencoders Shuiwang Ji Department of Computer Science &amp; Engineering Texas A&amp;M University 1 / 25 Orthogonal Matrices 1 An orthogonal matrix is a square matrix whose columns and rows are orthogonal unit

419 views • 25 slides
Periodically weighted tilings and (matrix) Maurice Duits Royal Institute of Technology

Periodically weighted tilings and (matrix) Maurice Duits Royal Institute of Technology

May 9, 2019 Seoul Periodically weighted tilings and (matrix) Maurice Duits Royal Institute of Technology orthogonal polynomials Based on joint works with : A.B.J. Kuijlaars, The two periodic Aztec diamond and Matrix orthogonal

438 views • 43 slides
Sum of matrix entries of representations of the symmetric group and its asymptotics Dario De

Sum of matrix entries of representations of the symmetric group and its asymptotics Dario De

Sum of matrix entries of representations of the symmetric group and its asymptotics Sum of matrix entries of representations of the symmetric group and its asymptotics Dario De Stavola 13 October 2015 Advisor: Valentin Fray Affiliation:

851 views • 68 slides
Unitary Matrix Models: universality conjecture in the bulk and orthogonal and on the edge of the

Unitary Matrix Models: universality conjecture in the bulk and orthogonal and on the edge of the

Unitary Matrix Models Poplavskyi M. Random matrices Unitary Matrix Models: universality conjecture in the bulk and orthogonal and on the edge of the spectrum. polynomials Overview of previous results Global M. Poplavskyi regime

354 views • 22 slides
New examples of matrix orthogonal polynomials satisfying second order differential equations.

New examples of matrix orthogonal polynomials satisfying second order differential equations.

Introduction A new family Rodrigues formula Recurrence relations New examples of matrix orthogonal polynomials satisfying second order differential equations. Structural formulas J. Borrego Morell ( ) , joint work with A. Dur an (

650 views • 52 slides
Computing the sets of totally symmetric and totally conjugate orthogonal partial Latin squares by

Computing the sets of totally symmetric and totally conjugate orthogonal partial Latin squares by

Computing the sets of totally symmetric and totally conjugate orthogonal partial Latin squares by means of a SAT solver Ra ul M. Falc on Department of Applied Mathematics I. University of Seville. (Joint work with Oscar J. Falc on

521 views • 36 slides
Greedy Orthogonal Pivoting for Non-negative Matrix Factorization Kai Zhang, Jun Liu, Jie Zhang,

Greedy Orthogonal Pivoting for Non-negative Matrix Factorization Kai Zhang, Jun Liu, Jie Zhang,

Greedy Orthogonal Pivoting for Non-negative Matrix Factorization Kai Zhang, Jun Liu, Jie Zhang, Jun Wang Infinia ML Inc., Fudan University, East China Normal University Non-negative Matrix Factorization Represent data with non-negative basis

330 views • 7 slides
Multiple Orthogonal Polynomials and the Normal Matrix Model Arno Kuijlaars Department of

Multiple Orthogonal Polynomials and the Normal Matrix Model Arno Kuijlaars Department of

Multiple Orthogonal Polynomials and the Normal Matrix Model Arno Kuijlaars Department of Mathematics KU Leuven, Belgium Joint work with Pavel Bleher Adv. Math. 2012 Random Matrices and their Applications T el ecom ParisTech, Paris, 8

1.15k views • 74 slides
Orthogonal functions with a skew-Hermitian differentiation matrix Adhemar Bultheel Dept.

Orthogonal functions with a skew-Hermitian differentiation matrix Adhemar Bultheel Dept.

Orthogonal functions with a skew-Hermitian differentiation matrix Adhemar Bultheel Dept. Computer Science, KU Leuven ETNA25, Recent Advances in Scientific Computation Santa Margherita di Pula, Sardinia, May 27-30, 2019

663 views • 43 slides
Asymptotics of orthogonal polynomials in normal matrix ensemble Seung-Yeop Lee (University of

Asymptotics of orthogonal polynomials in normal matrix ensemble Seung-Yeop Lee (University of

Asymptotics of orthogonal polynomials in normal matrix ensemble Seung-Yeop Lee (University of South Florida) Cincinnati, September 20th 2014 1 / 32 Joint work with Roman Riser. Many discussions with Marco Bertola, Robert Buckingham, Maurice

918 views • 32 slides
New examples of matrix orthogonal polynomials satisfying second order differential equations.

New examples of matrix orthogonal polynomials satisfying second order differential equations.

New examples of matrix orthogonal polynomials satisfying second order differential equations. Structural formulas Joint work with A.J. Dur an University of Doc-course,University of Seville, spring 2010 Introduction Given a self-adjoint

424 views • 9 slides
Symmetric Pattern Based Word Embeddings for Improved Word Similarity Prediction Roy Schwartz + ,

Symmetric Pattern Based Word Embeddings for Improved Word Similarity Prediction Roy Schwartz + ,

Symmetric Pattern Based Word Embeddings for Improved Word Similarity Prediction Roy Schwartz + , Roi Reichart * and Ari Rappoport + + The Hebrew University, * Technion IIT CoNLL 2015 Symmetric Pattern Based Word Embeddings for 2 Improved Word

917 views • 49 slides
6. The Symmetric Eigenvalue Problem A must for engineers . . . 6. The Symmetric Eigenvalue

6. The Symmetric Eigenvalue Problem A must for engineers . . . 6. The Symmetric Eigenvalue

Motivation Condition Vector Iteration QR Iteration Reduction Algorithms 6. The Symmetric Eigenvalue Problem A must for engineers . . . 6. The Symmetric Eigenvalue Problem Numerical Programming I (for CSE), Hans-Joachim Bungartz page 1

430 views • 28 slides
Linear Algebra Chapter 10: Solving Large Systems Section 10.2 The LU -FactorizationProofs of

Linear Algebra Chapter 10: Solving Large Systems Section 10.2 The LU -FactorizationProofs of

Linear Algebra Chapter 10: Solving Large Systems Section 10.2 The LU -FactorizationProofs of Theorems July 5, 2020 () Linear Algebra July 5, 2020 1 / 6 Table of contents Theorem 10.A 1 Theorem 10.1. Unique Factorization 2 () Linear

244 views • 11 slides
Efficient Full-Matrix Adaptive Regularization Naman Agarwal, Brian Bullins, Xinyi Chen, Elad

Efficient Full-Matrix Adaptive Regularization Naman Agarwal, Brian Bullins, Xinyi Chen, Elad

Efficient Full-Matrix Adaptive Regularization Naman Agarwal, Brian Bullins, Xinyi Chen, Elad Hazan, Karan Singh, Cyril Zhang, Yi Zhang Princeton University Google AI Princeton Adaptive Preconditioning in ML Optimization in ML: training

344 views • 11 slides
Accuracy and Stability: recent advances in C.A.G.D. J.M. Pe na* *University of Zaragoza, Spain

Accuracy and Stability: recent advances in C.A.G.D. J.M. Pe na* *University of Zaragoza, Spain

Accuracy and Stability: recent advances in C.A.G.D. J.M. Pe na* *University of Zaragoza, Spain M.A.I.A. 2013 , Erice 1 Effects of finite precision arithmetic on numerical algorithms: Roundoff errors. Data uncertainty. Key concepts:

485 views • 44 slides
On the Linear Algebra Employed in the MOSEK Conic Optimizer Monday Jul 13 Erling D. Andersen

On the Linear Algebra Employed in the MOSEK Conic Optimizer Monday Jul 13 Erling D. Andersen

On the Linear Algebra Employed in the MOSEK Conic Optimizer Monday Jul 13 Erling D. Andersen www.mosek.com MOSEK summary General LP Convex QP SDP MIP CQP Version 8: Work in progress. The conic optimization problem n n

564 views • 22 slides
s r t st tr

s r t st tr

s r t st tr r st Prr s

730 views • 20 slides
Matrix Calculations: Diagonalisation, Orthogonality, and Applications A. Kissinger Institute for

Matrix Calculations: Diagonalisation, Orthogonality, and Applications A. Kissinger Institute for

Eigenvectors and diagonalisation Inner products and orthogonality Radboud University Nijmegen Wrapping up Matrix Calculations: Diagonalisation, Orthogonality, and Applications A. Kissinger Institute for Computing and Information Sciences

758 views • 53 slides
Lecture 12: Matrices Dr. Chengjiang Long Computer Vision Researcher at Kitware Inc. Adjunct

Lecture 12: Matrices Dr. Chengjiang Long Computer Vision Researcher at Kitware Inc. Adjunct

Lecture 12: Matrices Dr. Chengjiang Long Computer Vision Researcher at Kitware Inc. Adjunct Professor at SUNY at Albany. Email: clong2@albany.edu Important timeline Midterm exam 1 Final exam Midterm exam 2 Midterm exam 1: Oct 8th, 2018

811 views • 36 slides
Linear Algebra Linear algebra has become as basic and as applicable as calculus, and

Linear Algebra Linear algebra has become as basic and as applicable as calculus, and

Mathematical Tools for Neural and Cognitive Science Fall semester, 2018 Section 1: Linear Algebra Linear Algebra Linear algebra has become as basic and as applicable as calculus, and fortunately it is easier - Gilbert Strang, Linear

359 views • 10 slides

More recommend