SlideShare uma empresa Scribd logo
1 de 64
Baixar para ler offline
82nd GAMM Annual Scientific Conference
                                                        Graz, April 19, 2011




                   Preconditioned Inverse Iteration for
                         Hierarchical Matrices
                                 Peter Benner and Thomas Mach

                   Max Planck Institute for Dynamics of Complex Technical Systems
                       Computational Methods in Systems and Control Theory
                                             Magdeburg




                                                                                                           MAX PLANCK INSTITUTE
                                                                                                         FOR DYNAMICS OF COMPLEX
                                                                                                            TECHNICAL SYSTEMS
                                                                                                                MAGDEBURG




Max Planck Institute Magdeburg         Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices        1/28
Hierarchical (H-)Matrices                      PINVIT                                        Numerical Results



   Outline



        1   Hierarchical (H-)Matrices

        2   PINVIT

        3   Numerical Results




Max Planck Institute Magdeburg   Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   2/28
Hierarchical (H-)Matrices                                   PINVIT                                        Numerical Results



   H-Matrices                    [Hackbusch 1998]


        Some dense matrices, e.g. BEM or FEM, can be approximated by
        H-matrices in a data-sparse manner.


        hierarchical tree TI                               block H-tree TI × I
         I = {1, 2, 3, 4, 5, 6, 7, 8}
                                            12345678            12345678            12345678            12345678
                                        1                   1                   1                   1
                                        2                   2                   2                   2
         {1, 2, 3, 4}   {5, 6, 7, 8}    3                   3                   3                   3
                                        4                   4                   4                   4
                                        5                   5                   5                   5
                                        6                   6                   6                   6
        {1, 2} {3, 4} {5, 6} {7, 8}     7                   7                   7                   7
                                        8                   8                   8                   8


       {1}{2}{3}{4}{5}{6}{7}{8}                  dense matrices, rank-k-matrices

        rank-k-matrix: Ms×t = AB T , A ∈ Rn×k , B ∈ Rm×k (k                                             n, m)


Max Planck Institute Magdeburg                Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   3/28
Hierarchical (H-)Matrices                                                                                                                                                                                                                                                                                                  PINVIT                                        Numerical Results



   H-Matrices                                                                                                                                                                        [Hackbusch 1998]


        Hierarchical matrices
         H(TI × I , k) = M ∈ RI × I rank (Ma×b ) ≤ k ∀a × b admissible
                     3       3




                22 7         10       14        8    11       9                                8           5
            3   7    19      10                                                            11      8


            3   10 10        31       11        11   9        16            12             8       15 8             12                                                        11 8
                                  19       10                                                                                                                            11      8

                14

                8
                11
                         11
                         11
                         8
                                  10


                                      11
                                           31 11
                                                31       11
                                                                        9
                                                                        3
                                                                            16 11
                                                                                 7
                                                                                 3
                                                                                      8
                                                                                                                6

                                                                                                                15
                                                                                                                         5
                                                                                                                         6
                                                                                                                                   7
                                                                                                                                        13
                                                                                                                                             5
                                                                                                                                                                         8



                                                                                                                                                                             8
                                                                                                                                                                                 15 9

                                                                                                                                                                                      15            12                13
                                                                                                                                                                                                                                                                                                                   adaptive rank k(ε)
                                                                                                                                                                                                                                                           19
                                                                                                                                   13
                                       11                61                                                                                               11                                                         10 7
                                                                                                                                        8

                9        16                                                 11        11                                           8    11   8




                                                                                                       13                                                                            13
                                                     9    3             25       10                                                                                                                              13   8

                                                     7        11        10       19   11                                                              6        5                                                 8    11   9

                     11 16
                                                          3




                                                                                                                                                                                                                                                                                                                   storage NSt,H (T , k) = O(n log n k(ε))
                                                     8        11            11        31                                               11             15       6                                                 9         15       11
                     11      8


                8    8       15                                                                                 10       10        15        9

                5        8             11                                                       61              3




                                                                                                                6        14        9         11           10                                                                                  10 7

                                                                  13
                                      6         15                                             10      3   6    25   10       6                                                                                                           13       8


                     12               5         6                                              10          14 6
                                                                                                                10   19


                                                                                                                     10
                                                                                                                              10


                                                                                                                              31       10                 16                                                                              9
                                                                                                                                                                                                                                               11



                                                                                                                                                                                                                                               8
                                                                                                                                                                                                                                                       8


                                                                                                                                                                                                                                                       15        11                   12

                                                                                                                                                                                          20
                                                              13   8




                                                                                                                                                                                                                                                                                                                   complexity of approximate arithmetic
                                                     7        8    11                          15          9                                          10       10
                                                     5        8             11                 8           11       10                 51             7    9
                                                                                                                                                           3
                                                                                                                                                                    7
                                                                                                                                                                    3
                                                                                                                                                                                                                                                                                     9             7
                                                                            6         15                                           10                 25       11                                                                                                                        8

                         13                                                                                                                                                                                                                        13
                                                                                                                                             7                                                                                                                                   13


                                                         11                 5         6         10 16                              10
                                                                                                                                             9

                                                                                                                                             7
                                                                                                                                                  3




                                                                                                                                                  3
                                                                                                                                                      11
                                                                                                                                                           25
                                                                                                                                                           10
                                                                                                                                                                    10

                                                                                                                                                                    19                                                                                                           9
                                                                                                                                                                                                                                                                                      13
                                                                                                                                                                                                                                                                                      8
                                                                                                                                                                                                                                                                                              8

                                                                                                                                                                                                                                                                                              11
                                                                                                                                                                                                                                                                                                   11
                                  11       8                                                                                                                                               3   7


                                  8        15   8                                                                                                                            39       10       10       10                                6        5
                     10
                                                                                                                                                                                           3




                                      9         15                                                                                                                       3
                                                                                                                                                                             10  3
                                                                                                                                                                                      25 7          10
                                                                                                                                                                                                    3
                                                                                                                                                                                                            6
                                                                                                                                                                                                            3
                                                                                                                                                                                                                     15 10                13       6             11

                                                                  13
                                                                                                                                                                         7       10   7        22 7         10                  11   9                          8           5


                                                                                                                                                                                                                                                                                      12
                                                                                                                                                                                                                                                                                                                                      O(n log n k(ε))
                                                                                                                                                                                      10       7    19      10                                              11      8




                                                                                                                                                                                                                                                                                                                   MH v
                                                                                                                                                                                           3




                     7                 12                                                                                                                                    10       6    3   10 10        31       11         9    16       12            8       15 8




                                                                                                               20
                                                                        13       8
                                                                        8        11   8                                                                                                                          34        10   13   10                                          6       5
                                                         10                 9         15                                                                                      15                    11           10        25   7    11                                          13      6         11

                         13                                                                                                                                                                                                                        13
                                                                                                                                                                                               11       8        13        7




                                                                                                                                                                                                                                                                                                                                      O(n log n k(ε)2 )
                                                         7                  11                                                                                                10                        16                          61                                               11 23

                                                                                                                                                                                                                                                                                                                   +H , −H
                                                                                                                                                                                               9                 10        11
                                                                                                                13       9                                                   6        13                                                  20       9            9           7
                                                                                                                     11       8                                                                                                                                         3   7

                                                                                                9               8    8        15                                             5        6             12                                    9        39 10                3   10       15 10

                                                                                                                                        12                                                                            13
                                                                                                                                                                                                    11      8


                                                                                                                                                                                               8    8       15                            9        10                                              15   9

                                                                                                7                   11                                                        11                                                               3       3




                                                                                                                                                                                                                                                                 61 10
                                                                                                                                                                                                                                                                                                                                   −1
                                                                                                                                                                                                                                                                                                                   ∗H , HLU(·), (·)H O(n (log n)2 k(ε)2 )
                                                                                                                                                                                               5        8                                 7    7       10                                          9    11




                                  19
                                                                                                                                                      13       9                                                 6         13                                                    20      9         9    7


                                                                                                                                       9              8
                                                                                                                                                           13
                                                                                                                                                           8
                                                                                                                                                                    8
                                                                                                                                                                    11                                           5         6        10        15                 10              9       34 10          13




                                                                                                       12                              7                  11                         12                              11             23        10
                                                                                                                                                                                                                                                                15
                                                                                                                                                                                                                                                                8
                                                                                                                                                                                                                                                                            9
                                                                                                                                                                                                                                                                            11
                                                                                                                                                                                                                                                                                 9
                                                                                                                                                                                                                                                                                 7
                                                                                                                                                                                                                                                                                         10
                                                                                                                                                                                                                                                                                         13        51




Max Planck Institute Magdeburg                                                                                                                                                                                                                                                                               Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   4/28
Hierarchical (H-)Matrices                                                                                                                                                                                                                                                                                                  PINVIT                                        Numerical Results



   H-Matrices                                                                                                                                                                        [Hackbusch 1998]


        Hierarchical matrices
         H(TI × I , k) = M ∈ RI × I rank (Ma×b ) ≤ k ∀a × b admissible
                     3       3




                22 7         10       14        8    11       9                                8           5
            3   7    19      10                                                            11      8


            3   10 10        31       11        11   9        16            12             8       15 8             12                                                        11 8
                                  19       10                                                                                                                            11      8

                14

                8
                11
                         11
                         11
                         8
                                  10


                                      11
                                           31 11
                                                31       11
                                                                        9
                                                                        3
                                                                            16 11
                                                                                 7
                                                                                 3
                                                                                      8
                                                                                                                6

                                                                                                                15
                                                                                                                         5
                                                                                                                         6
                                                                                                                                   7
                                                                                                                                        13
                                                                                                                                             5
                                                                                                                                                                         8



                                                                                                                                                                             8
                                                                                                                                                                                 15 9

                                                                                                                                                                                      15            12                13
                                                                                                                                                                                                                                                                                                                   adaptive rank k(ε)
                                                                                                                                                                                                                                                           19
                                                                                                                                   13
                                       11                61                                                                                               11                                                         10 7
                                                                                                                                        8

                9        16                                                 11        11                                           8    11   8




                                                                                                       13                                                                            13
                                                     9    3             25       10                                                                                                                              13   8

                                                     7        11        10       19   11                                                              6        5                                                 8    11   9

                     11 16
                                                          3




                                                                                                                                                                                                                                                                                                                   storage NSt,H (T , k) = O(n log n k(ε))
                                                     8        11            11        31                                               11             15       6                                                 9         15       11
                     11      8


                8    8       15                                                                                 10       10        15        9

                5        8             11                                                       61              3




                                                                                                                6        14        9         11           10                                                                                  10 7

                                                                  13
                                      6         15                                             10      3   6    25   10       6                                                                                                           13       8


                     12               5         6                                              10          14 6
                                                                                                                10   19


                                                                                                                     10
                                                                                                                              10


                                                                                                                              31       10                 16                                                                              9
                                                                                                                                                                                                                                               11



                                                                                                                                                                                                                                               8
                                                                                                                                                                                                                                                       8


                                                                                                                                                                                                                                                       15        11                   12

                                                                                                                                                                                          20
                                                              13   8




                                                                                                                                                                                                                                                                                                                   complexity of approximate arithmetic
                                                     7        8    11                          15          9                                          10       10
                                                     5        8             11                 8           11       10                 51             7    9
                                                                                                                                                           3
                                                                                                                                                                    7
                                                                                                                                                                    3
                                                                                                                                                                                                                                                                                     9             7
                                                                            6         15                                           10                 25       11                                                                                                                        8

                         13                                                                                                                                                                                                                        13
                                                                                                                                             7                                                                                                                                   13


                                                         11                 5         6         10 16                              10
                                                                                                                                             9

                                                                                                                                             7
                                                                                                                                                  3




                                                                                                                                                  3
                                                                                                                                                      11
                                                                                                                                                           25
                                                                                                                                                           10
                                                                                                                                                                    10

                                                                                                                                                                    19                                                                                                           9
                                                                                                                                                                                                                                                                                      13
                                                                                                                                                                                                                                                                                      8
                                                                                                                                                                                                                                                                                              8

                                                                                                                                                                                                                                                                                              11
                                                                                                                                                                                                                                                                                                   11
                                  11       8                                                                                                                                               3   7


                                  8        15   8                                                                                                                            39       10       10       10                                6        5
                     10
                                                                                                                                                                                           3




                                      9         15                                                                                                                       3
                                                                                                                                                                             10  3
                                                                                                                                                                                      25 7          10
                                                                                                                                                                                                    3
                                                                                                                                                                                                            6
                                                                                                                                                                                                            3
                                                                                                                                                                                                                     15 10                13       6             11

                                                                  13
                                                                                                                                                                         7       10   7        22 7         10                  11   9                          8           5


                                                                                                                                                                                                                                                                                      12
                                                                                                                                                                                                                                                                                                                                      O(n log n k(ε))
                                                                                                                                                                                      10       7    19      10                                              11      8




                                                                                                                                                                                                                                                                                                                   MH v
                                                                                                                                                                                           3




                     7                 12                                                                                                                                    10       6    3   10 10        31       11         9    16       12            8       15 8




                                                                                                               20
                                                                        13       8
                                                                        8        11   8                                                                                                                          34        10   13   10                                          6       5
                                                         10                 9         15                                                                                      15                    11           10        25   7    11                                          13      6         11

                         13                                                                                                                                                                                                                        13
                                                                                                                                                                                               11       8        13        7




                                                                                                                                                                                                                                                                                                                                      O(n log n k(ε)2 )
                                                         7                  11                                                                                                10                        16                          61                                               11 23

                                                                                                                                                                                                                                                                                                                   +H , −H
                                                                                                                                                                                               9                 10        11
                                                                                                                13       9                                                   6        13                                                  20       9            9           7
                                                                                                                     11       8                                                                                                                                         3   7

                                                                                                9               8    8        15                                             5        6             12                                    9        39 10                3   10       15 10

                                                                                                                                        12                                                                            13
                                                                                                                                                                                                    11      8


                                                                                                                                                                                               8    8       15                            9        10                                              15   9

                                                                                                7                   11                                                        11                                                               3       3




                                                                                                                                                                                                                                                                 61 10
                                                                                                                                                                                                                                                                                                                                   −1
                                                                                                                                                                                                                                                                                                                   ∗H , HLU(·), (·)H O(n (log n)2 k(ε)2 )
                                                                                                                                                                                               5        8                                 7    7       10                                          9    11




                                  19
                                                                                                                                                      13       9                                                 6         13                                                    20      9         9    7


                                                                                                                                       9              8
                                                                                                                                                           13
                                                                                                                                                           8
                                                                                                                                                                    8
                                                                                                                                                                    11                                           5         6        10        15                 10              9       34 10          13




                                                                                                       12                              7                  11                         12                              11             23        10
                                                                                                                                                                                                                                                                15
                                                                                                                                                                                                                                                                8
                                                                                                                                                                                                                                                                            9
                                                                                                                                                                                                                                                                            11
                                                                                                                                                                                                                                                                                 9
                                                                                                                                                                                                                                                                                 7
                                                                                                                                                                                                                                                                                         10
                                                                                                                                                                                                                                                                                         13        51




                                                                                       T
                                                                                      B1                                                                                                                                                                                                                          T
                                                                                                                                                                                                                                                                                                                 B2                                     T
                                                                                                                                                                                                                                                                                                                                                       B1
         A1                                                                                                                                                                                              + A2                                                                                                                 = A1 A2                   T
                                                                                                                                                                                                                                                                                                                                                       B2

Max Planck Institute Magdeburg                                                                                                                                                                                                                                                                               Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   4/28
Hierarchical (H-)Matrices                         PINVIT                                        Numerical Results



   Goal: Eigenvalues of symmetric H-Matrices


                               M = M T ∈ H(T , k),
                             Λ(M) = {λ1, λ2, . . . , λn },
                             λ1 ≥ λ2 ≥ · · · ≥ λn > 0
                                                ⇓
               How to find λi in O(n (log n)α k β )?




Max Planck Institute Magdeburg      Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   5/28
Hierarchical (H-)Matrices                         PINVIT                                        Numerical Results



   Goal: Eigenvalues of symmetric H-Matrices


                               M = M T ∈ H(T , k),
                             Λ(M) = {λ1, λ2, . . . , λn },
                             λ1 ≥ λ2 ≥ · · · ≥ λn > 0
                                                ⇓
               How to find λn in O(n (log n)α k β )?




Max Planck Institute Magdeburg      Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   5/28
Hierarchical (H-)Matrices                         PINVIT                                        Numerical Results



   Preconditioned Inverse Iteration
   [Knyazev, Neymeyr, et al.]


        Definition
        The function
                                                             x T Mx
                                 µ(x) = µ(x, M) =
                                                              xT x
        is called the Rayleigh quotient.




Max Planck Institute Magdeburg      Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   6/28
Hierarchical (H-)Matrices                         PINVIT                                        Numerical Results



   Preconditioned Inverse Iteration
   [Knyazev, Neymeyr, et al.]


        Definition
        The function
                                                             x T Mx
                                 µ(x) = µ(x, M) =
                                                              xT x
        is called the Rayleigh quotient.


        Minimize the Rayleigh quotient by a gradient method:
                                                 2
             xi+1 := xi − α µ(xi ),     µ(x) = T (Mx − xµ(x)) ,
                                                x x




Max Planck Institute Magdeburg      Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   6/28
Hierarchical (H-)Matrices                         PINVIT                                        Numerical Results



   Preconditioned Inverse Iteration
   [Knyazev, Neymeyr, et al.]




                                 Residual r (x) = Mx − xµ(x).




        Minimize the Rayleigh quotient by a gradient method:
                                                 2
             xi+1 := xi − α µ(xi ),     µ(x) = T (Mx − xµ(x)) ,
                                                x x




Max Planck Institute Magdeburg      Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   6/28
Hierarchical (H-)Matrices                              PINVIT                                        Numerical Results



   Preconditioned Inverse Iteration
   [Knyazev, Neymeyr, et al.]


        Definition
        The function
                                                                  x T Mx
                                     µ(x) = µ(x, M) =
                                                                   xT x
        is called the Rayleigh quotient.


        Minimize the Rayleigh quotient by a gradient method:
                                                 2
             xi+1 := xi − α µ(xi ),     µ(x) = T (Mx − xµ(x)) ,
                                                x x
        + preconditioning ⇒ update equation:
                                 xi+1 := xi − B −1 (Mxi − xi µ(xi )) .

Max Planck Institute Magdeburg           Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   6/28
Hierarchical (H-)Matrices                              PINVIT                                        Numerical Results



   Preconditioned Inverse Iteration
   [Knyazev, Neymeyr, et al.]




          Preconditioned residual B −1 r (x) = B −1 (Mx − xµ(x)).




        + preconditioning ⇒ update equation:
                                 xi+1 := xi − B −1 (Mxi − xi µ(xi )) .

Max Planck Institute Magdeburg           Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   6/28
Hierarchical (H-)Matrices                             PINVIT                                        Numerical Results



   Preconditioned Inverse Iteration
   [Knyazev, Neymeyr 2009]


                                 xi+1 := xi − B −1 (Mxi − xi µ(xi ))


        If
                M ∈ Rn×n symmetric positive definite and
                B −1 approximates the inverse of M, so that

                                         I − B −1 M          M
                                                                 ≤ c < 1,

        then Preconditioned INVerse ITeration (PINVIT) converges and
        the number of iterations is independent of n.




Max Planck Institute Magdeburg          Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   7/28
Hierarchical (H-)Matrices                       PINVIT                                        Numerical Results



   Preconditioned Inverse Iteration
   [Knyazev, Neymeyr, et al.]

        The residual

                                  ri = Mxi − xi µ(xi )

        converges to 0, so that

                                           ri   2   <

        is a useful termination criterion.




Max Planck Institute Magdeburg    Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   8/28
Hierarchical (H-)Matrices                           PINVIT                                        Numerical Results



   Variants of PINVIT
   [Neymeyr 2001: A Hierarchy of Precond. Eigens. for Ellipt. Diff. Op.]


        Classification by Neymeyr:

                PINVIT(1): xi+1 := xi − B −1 ri .
                PINVIT(2): xi+1 := arg minv ∈span{xi ,B −1 ri } µ(v ).
                PINVIT(3): xi+1 := arg minv ∈span{xi−1 ,xi ,B −1 ri } µ(v ).
                PINVIT(n): Analogously.
                PINVIT(·,d): Replacing x by a rectangular full rank matrix
                X ∈ Rn×d one gets the subspace version of PINVIT(·).




Max Planck Institute Magdeburg        Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   9/28
Hierarchical (H-)Matrices                      PINVIT                                        Numerical Results



   Algorithm and Complexity
        The number of iterations is independent of matrix size n.

        H-PINVIT(1,d)
        Input: M ∈ Rn×n , X0 ∈ Rn×d (X0 X0 = I , e.g. randomly chosen)
                                        T

        Output: Xp ∈ R n×d , µ ∈ Rd×d , with MX − X µ ≤
                                                p      p

        B −1 = (M)−1 or B −1 = L−T L−1
                    H            H   H
                                  T
        R := MX0 − X0 µ, µ = X0 MX0
        for (i := 1; R F > ; i + +) do
            Xi := Orthogonalize Xi−1 − B −1 R
            R := MXi − Xi µ, µ = XiT MXi
        end




Max Planck Institute Magdeburg   Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   10/28
Hierarchical (H-)Matrices                      PINVIT                                        Numerical Results



   Algorithm and Complexity
        The number of iterations is independent of matrix size n.

        H-PINVIT(1,d)
        Input: M ∈ Rn×n , X0 ∈ Rn×d (X0 X0 = I , e.g. randomly chosen)
                                        T

        Output: Xp ∈ R n×d , µ ∈ Rd×d , with MX − X µ ≤
                                                p      p

        B −1 = (M)−1 or B −1 = L−T L−1
                    H            H   H                                 O(n (log n)2 k (c)2 )
                                  T
        R := MX0 − X0 µ, µ = X0 MX0
        for (i := 1; R F > ; i + +) do
            Xi := Orthogonalize Xi−1 − B −1 R                            O(n (log n) k (c)2 )
            R := MXi − Xi µ, µ = XiT MXi                                  O(n (log n) k (c))
        end

        The complexity of the algorithm is determined by the H-matrix
        inversion/Cholesky decomposition: ⇒ O(n (log n)2 k (c)2 ).

Max Planck Institute Magdeburg   Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   10/28
Hierarchical (H-)Matrices                         PINVIT                                        Numerical Results



   Goal: Eigenvalues of symmetric H-Matrices


                               M = M T ∈ H(T , k),
                             Λ(M) = {λ1, λ2, . . . , λn },
                             λ1 ≥ λ2 ≥ · · · ≥ λn > 0
                                                ⇓
               How to find λn in O(n (log n)α k β )?




Max Planck Institute Magdeburg      Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   11/28
Hierarchical (H-)Matrices                         PINVIT                                        Numerical Results



   Goal: Eigenvalues of symmetric H-Matrices


                               M = M T ∈ H(T , k),
                             Λ(M) = {λ1, λ2, . . . , λn },
                             λ1 ≥ λ2 ≥ · · · ≥ λn > 0
                                                ⇓
               How to find λn in O(n (log n)α k β )?




Max Planck Institute Magdeburg      Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   11/28
Hierarchical (H-)Matrices                         PINVIT                                        Numerical Results



   Goal: Eigenvalues of symmetric H-Matrices


                               M = M T ∈ H(T , k),
                             Λ(M) = {λ1, λ2, . . . , λn },
                             λ1 ≥ λ2 ≥ · · · ≥ λn > 0
                                                ⇓
               How to find λn in O(n (log n)α k β )?

               How to find λi in O(n (log n)α k β )?


Max Planck Institute Magdeburg      Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   12/28
Hierarchical (H-)Matrices                      PINVIT                                        Numerical Results



   Folded Spectrum Method

                   How to find λi in O(n (log n)α k β )?
                    If i = n − d with d < O(log n),
                   use subspace version PINVIT(·,d ).


                                                     ...
                 0λn λn−1 λn−2                                                            λ1




Max Planck Institute Magdeburg   Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   13/28
Hierarchical (H-)Matrices                            PINVIT                                        Numerical Results



   Folded Spectrum Method

                   How to find λi in O(n (log n)α k β )?
                     If i = n − d with d    log n?



                                 ...                                          ...
                 0λn                   λi+1 λi            λi−1                                  λ1




Max Planck Institute Magdeburg         Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   13/28
Hierarchical (H-)Matrices                            PINVIT                                        Numerical Results



   Folded Spectrum Method

                   How to find λi in O(n (log n)α k β )?
                     If i = n − d with d      log n,
                           shift with σ near λi .


                                 ...                                          ...
                 0λn                   λi+1 λi            λi−1                                  λ1
                                             σ




Max Planck Institute Magdeburg         Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   13/28
Hierarchical (H-)Matrices                            PINVIT                                        Numerical Results



   Folded Spectrum Method

                   How to find λi in O(n (log n)α k β )?
                     If i = n − d with d      log n,
                           shift with σ near λi .


                                 ...                                          ...
                 0λn                   λi+1 λi            λi−1                                  λ1
                                             σ



                But (M − σI) is not positive definite.
Max Planck Institute Magdeburg         Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   13/28
Hierarchical (H-)Matrices                            PINVIT                                        Numerical Results



   Folded Spectrum Method

        Folded Spectrum Method                                             [Wang, Zunger 1994]


                                      Mσ = (M − σI)2

        Mσ is s.p.d., if M is s.p.d. and σ = λi .
        Assume all eigenvalues of Mσ are simple.

                             Mv = λv ⇔ Mσ v = (M − σI)2 v
                                                  = M 2 v − 2σMv + σ 2 v
                                                  = λ2 v − 2σλv + σ 2 v
                                                  = (λ − σ)2 v


Max Planck Institute Magdeburg         Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   14/28
Hierarchical (H-)Matrices                       PINVIT                                        Numerical Results



   Folded Spectrum Method
           1    Choose σ.
           2    Compute
                                  2
                  a) Mσ = (M − σI) and
                      −1
                  b) Mσ or LLT = Mσ .
           3    Use PINVIT to find the smallest eigenpair (µσ , v ) of Mσ .
           4    Compute µ = v T Mv /v T v .
        (µ, v ) is the nearest eigenpair to σ.




Max Planck Institute Magdeburg    Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   15/28
Hierarchical (H-)Matrices                       PINVIT                                        Numerical Results



   Folded Spectrum Method
           1    Choose σ.
           2    Compute
                                  2
                  a) Mσ = (M − σI) and
                      −1
                  b) Mσ or LLT = Mσ .
           3    Use PINVIT to find the smallest eigenpair (µσ , v ) of Mσ .
           4    Compute µ = v T Mv /v T v .
        (µ, v ) is the nearest eigenpair to σ.

        H-arithmetic enables us to shift, square and invert M resp.
        Mσ in linear-polylogarithmic complexity.




Max Planck Institute Magdeburg    Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   15/28
Hierarchical (H-)Matrices                       PINVIT                                        Numerical Results



   Folded Spectrum Method
           1    Choose σ.
           2    Compute
                                  2
                  a) Mσ = (M − σI) and
                      −1
                  b) Mσ or LLT = Mσ .
           3    Use PINVIT to find the smallest eigenpair (µσ , v ) of Mσ .
           4    Compute µ = v T Mv /v T v .
        (µ, v ) is the nearest eigenpair to σ.

        H-arithmetic enables us to shift, square and invert M resp.
        Mσ in linear-polylogarithmic complexity.

        If M is sparse, then shifting, squaring and inverting is
        prohibitive.
Max Planck Institute Magdeburg    Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   15/28
Hierarchical (H-)Matrices                       PINVIT                                        Numerical Results



   Folded Spectrum Method
           1    Choose σ.
           2    Compute
                                  2
                  a) Mσ = (M − σI) and
                      −1
                  b) Mσ or LLT = Mσ .
           3    Use PINVIT to find the smallest eigenpair (µσ , v ) of Mσ .
           4    Compute µ = v T Mv /v T v .
        (µ, v ) is the nearest eigenpair to σ.

        H-arithmetic enables us to shift, square and invert M resp.
        Mσ in linear-polylogarithmic complexity.

        If M is sparse, then shifting, squaring and inverting is
        prohibitive.
Max Planck Institute Magdeburg    Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   15/28
Hierarchical (H-)Matrices                       PINVIT                                        Numerical Results



   Folded Spectrum Method
           1    Choose σ.
           2    Compute
                                  2
                  a) Mσ = (M − σI) and
                      −1
                  b) Mσ or LLT = Mσ .
           3    Use PINVIT to find the smallest eigenpair (µσ , v ) of Mσ .
           4    Compute µ = v T Mv /v T v .
        (µ, v ) is the nearest eigenpair to σ.

        H-arithmetic enables us to shift, square and invert M resp.
        Mσ in linear-polylogarithmic complexity.

        If M is sparse, then shifting, squaring and inverting is
        prohibitive.
Max Planck Institute Magdeburg    Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   15/28
Hierarchical (H-)Matrices                         PINVIT                                        Numerical Results



   Goal: Eigenvalues of symmetric H-Matrices


                               M = M T ∈ H(T , k),
                             Λ(M) = {λ1, λ2, . . . , λn },
                             λ1 ≥ λ2 ≥ · · · ≥ λn > 0
                                                ⇓
               How to find λn in O(n (log n)α k β )?

               How to find λi in O(n (log n)α k β )?


Max Planck Institute Magdeburg      Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   16/28
Hierarchical (H-)Matrices                         PINVIT                                        Numerical Results



   Goal: Eigenvalues of symmetric H-Matrices


                               M = M T ∈ H(T , k),
                             Λ(M) = {λ1, λ2, . . . , λn },
                             λ1 ≥ λ2 ≥ · · · ≥ λn > 0
                                                ⇓
               How to find λn in O(n (log n)α k β )?

               How to find λi in O(n (log n)α k β )?


Max Planck Institute Magdeburg      Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   16/28
Hierarchical (H-)Matrices                        PINVIT                                        Numerical Results



   Numerical Results




                                 Numerical Results




Max Planck Institute Magdeburg     Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   17/28
Hierarchical (H-)Matrices                      PINVIT                                        Numerical Results



   Hlib




        Hlib                                       ¨
                                                 [Borm, Grasedyck, et al.]
        We use the Hlib1.3 (www.hlib.org) for the
        H-arithmetic operations and some examples out of
        the library for testing the eigenvalue algorithm.




Max Planck Institute Magdeburg   Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   18/28
Hierarchical (H-)Matrices                                PINVIT                                        Numerical Results



   2D Laplace, smallest eigenvalues
        2D Laplace over [−1, 1] × [−1, 1], precond.: H-inversion

                                                                                       ti         N(ni )
              Name                    ni        error                     ti         ti−1        N(ni−1 )
              FEM8                    64    5.6146E-010                0.01
              FEM16                  256    4.5918E-010                0.02          2.00        106.67
              FEM32                1 024    3.7550E-010                0.12          6.17         27.08
              FEM64                4 096    3.8009E-010                0.82          6.68          8.06
              FEM128              16 384    4.4099E-010                5.84          7.09          5.44
              FEM256              65 536    3.9651E-010               34.47          5.91          5.22
              FEM512             262 144    3.7877E-010              194.00          5.63          6.51

        d = 4, c = 0.2, eig = 10−4 , N(ni ) = ni (log2 ni ) Csp Cid
        Time t only H-PINVIT (without H-inversion)


Max Planck Institute Magdeburg             Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   19/28
Hierarchical (H-)Matrices                                PINVIT                                        Numerical Results



   2D Laplace, smallest eigenvalues
        2D Laplace over [−1, 1] × [−1, 1], precond.: H-inversion

                                                                               ti    N(ni )
              Name                    ni        error                     ti ti−1   N(ni−1 )
              FEM8                    64    5.6146E-010                0.01
              FEM16                  256    4.5918E-010                0.02 2.00 106.67
              FEM32                1 024    3.7550E-010                0.12 6.17
                                                                                  27.08
              FEM64                4 096    3.8009E-010                        ˆ
                                                                       0.821 − λ1
                                                                          λ 6.68      8.06
              FEM128              16 384    4.4099E-010
                                                                              ˆ
                                                                          λ2 − λ2 
                                                                       5.84 7.09
                                                                                     5.44
              FEM256              65 536    3.9651E-010                 λ 5.91
                                                                              ˆ
                                                                      34.473 − λ3    5.22
              FEM512             262 144    3.7877E-010                        ˆ4
                                                                     194.004 − λ
                                                                          λ 5.63 2 6.51

        d = 4, c = 0.2, eig = 10−4 , N(ni ) = ni (log2 ni ) Csp Cid
        Time t only H-PINVIT (without H-inversion)


Max Planck Institute Magdeburg             Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   19/28
Hierarchical (H-)Matrices                                 PINVIT                                        Numerical Results



   2D Laplace, smallest eigenvalues
                          104
                                    H-PINVIT(1,4)               H-inversion
                          103       H-PINVIT(3,4)              O(N(ni ) log ni )
                                      O(N(ni ))
                          102
         CPU time in s




                          101

                          100

                         10−1

                         10−2

                         10−3
                               16      32        64     128 EM256 EM512
                           FEM      FEM      FEM    FEM     F     F
Max Planck Institute Magdeburg              Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   20/28
Hierarchical (H-)Matrices                                 PINVIT                                        Numerical Results



   2D Laplace, smallest eigenvalues
                          104
                                    H-PINVIT(1,4)               H-inversion
                          103       H-PINVIT(3,4)              O(N(ni ) log ni )
                                      O(N(ni ))                MATLAB eigs
                          102
         CPU time in s




                          101

                          100

                         10−1

                         10−2

                         10−3
                               16      32        64     128 EM256 EM512
                           FEM      FEM      FEM    FEM     F     F
Max Planck Institute Magdeburg              Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   20/28
Hierarchical (H-)Matrices                      PINVIT                                        Numerical Results



   Algorithm and Complexity


        H-PINVIT(1,d)
        Input: M ∈ Rn×n , X0 ∈ Rn×d (X0 X0 = I , e.g. randomly chosen)
                                        T

        Output: Xp ∈ R n×d , µ ∈ Rd×d , with MX − X µ ≤
                                                p      p

        B −1 = (M)−1 or B −1 = L−T L−1
                    H            H   H                                 O(n (log n)2 k (c)2 )
                                  T
        R := MX0 − X0 µ, µ = X0 MX0
        for (i := 1; R F > ; i + +) do
            Xi := Orthogonalize Xi−1 − B −1 R                            O(n (log n) k (c)2 )
            R := MXi − Xi µ, µ = XiT MXi                                  O(n (log n) k (c))
        end




Max Planck Institute Magdeburg   Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   21/28
Hierarchical (H-)Matrices                      PINVIT                                        Numerical Results



   Algorithm and Complexity


        H-PINVIT(1,d)
        Input: MCompetitive to Rn×d (X0 X0 = I , e.g. randomly chosen)
                 ∈ Rn×n , X0 ∈ MATLAB eigs.
                                        T

        Output: Xp ∈ R n×d , µ ∈ Rd×d , with MX − X µ ≤
                                                p      p

        B −1 = (M)−1 or B −1 = L−T L−1
                    H            H   H                                 O(n (log n)2 k (c)2 )
                                  T
        R := MX0 − X0 µ, µ = X0 MX0
        for (i := 1; R F > ; i + +) do
            Xi := Orthogonalize Xi−1 − B −1 R                            O(n (log n) k (c)2 )
            R := MXi − Xi µ, µ = XiT MXi                                  O(n (log n) k (c))
        end




Max Planck Institute Magdeburg   Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   21/28
Hierarchical (H-)Matrices                          PINVIT                                        Numerical Results



   Algorithm and Complexity


        H-PINVIT(1,d)
        Input: M ∈ Rn×n , X0 ∈ Rn×d (X0 X0 = I , e.g. randomly chosen)
                                        T

        Output: Xp ∈ R n×d , µ ∈ Rd×d , with MX − X µ ≤
                                                p      p

        B −1 = (M)−1 or B −1 = L−T L−1
                    H            H   H                                     O(n (log n)2 k (c)2 )
                                  T
        R := MX0 − X0 µ, µ = X0 MX0
        for (i := 1; R F > ; i + +) do
            Xi := Orthogonalize Xi−1 − B −1 R                                O(n (log n) k (c)2 )
            R := MXi − Xi µ, µ = XiT MXi                                      O(n (log n) k (c))
        end
                                 Expensive.



Max Planck Institute Magdeburg       Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   21/28
Hierarchical (H-)Matrices                                PINVIT                                        Numerical Results



   2D Laplace, smallest eigenvalues
        2D Laplace over [−1, 1] × [−1, 1], precond.: H-Cholesky decomp.

                                                                                       ti         N(ni )
              Name                    ni        error                      ti        ti−1        N(ni−1 )
              FEM8                    64    4.6920E-010                 0.01
              FEM16                  256    4.7963E-010                 0.02         2.00        106.67
              FEM32                1 024    3.4696E-010                 0.08         4.00         27.08
              FEM64                4 096    4.6414E-010                 0.48         6.00          8.06
              FEM128              16 384    3.3206E-010                 3.20         6.67          5.44
              FEM256              65 536    3.8468E-010                13.90         4.34          5.22
              FEM512             262 144    3.1353E-010                62.40         4.49          6.51

        d = 4, c = 0.2, eig = 10−4 , N(ni ) = ni (log2 ni ) Csp Cid
        Time t only H-PINVIT (without H-Cholesky decomposition)


Max Planck Institute Magdeburg             Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   22/28
Hierarchical (H-)Matrices                                 PINVIT                                        Numerical Results



   2D Laplace, smallest eigenvalues
                          104
                                    H-PINVIT(1,4)              H-Chol. decomp.
                          103       H-PINVIT(3,4)               O(N(ni ) log ni )
                                      O(N(ni ))
                          102
         CPU time in s




                          101

                          100

                         10−1

                         10−2

                         10−3
                               16      32        64     128 EM256 EM512
                           FEM      FEM      FEM    FEM     F     F
Max Planck Institute Magdeburg              Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   23/28
Hierarchical (H-)Matrices                                 PINVIT                                        Numerical Results



   2D Laplace, smallest eigenvalues
                          104
                                    H-PINVIT(1,4)              H-Chol. decomp.
                          103       H-PINVIT(3,4)               O(N(ni ) log ni )
                                      O(N(ni ))                 MATLAB eigs
                          102
         CPU time in s




                          101

                          100

                         10−1

                         10−2

                         10−3
                               16      32        64     128 EM256 EM512
                           FEM      FEM      FEM    FEM     F     F
Max Planck Institute Magdeburg              Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   23/28
Hierarchical (H-)Matrices                          PINVIT                                        Numerical Results



   2D Laplace, vn , FEM64
                                                                                   ·10−2
                                                                                       6

                                                                                         4
        ·10−2

            5                                                                            2

                                                                                         0
            0
                                                                                         −2
                                                                            1
         −5                                                                              −4
          −1                                                     0
                      −0.5       0
                                     0.5                                                 −6
                                                  1 −1


Max Planck Institute Magdeburg       Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   24/28
Hierarchical (H-)Matrices                          PINVIT                                        Numerical Results



   2D Laplace, vn−1 , FEM64
                                                                                   ·10−2
                                                                                       6

                                                                                         4
        ·10−2

            5                                                                            2

                                                                                         0
            0
                                                                                         −2
                                                                            1
         −5                                                                              −4
          −1                                                     0
                      −0.5       0
                                     0.5                                                 −6
                                                  1 −1


Max Planck Institute Magdeburg       Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   24/28
Hierarchical (H-)Matrices                          PINVIT                                        Numerical Results



   2D Laplace, vn−2 , FEM64
                                                                                   ·10−2
                                                                                       6

                                                                                         4
        ·10−2

            5                                                                            2

                                                                                         0
            0
                                                                                         −2
                                                                            1
         −5                                                                              −4
          −1                                                     0
                      −0.5       0
                                     0.5                                                 −6
                                                  1 −1


Max Planck Institute Magdeburg       Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   24/28
Hierarchical (H-)Matrices                          PINVIT                                        Numerical Results



   2D Laplace, vn−3 , FEM64
                                                                                   ·10−2
                                                                                       6

                                                                                         4
        ·10−2

            5                                                                            2

                                                                                         0
            0
                                                                                         −2
                                                                            1
         −5                                                                              −4
          −1                                                     0
                      −0.5       0
                                     0.5                                                 −6
                                                  1 −1


Max Planck Institute Magdeburg       Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   24/28
Hierarchical (H-)Matrices                               PINVIT                                        Numerical Results



   2D Laplace, inner eigenvalues
        2D Laplace over [−1, 1] × [−1, 1], precond.: H-inversion

                                                                                      ti          N(ni )
             Name                    ni       error                     ti          ti−1         N(ni−1 )
             FEM8                    64   1.7219E-013               <0.01
             FEM16                  256   7.2388E-012                0.03                        106.67
             FEM32                1 024   6.0045E-013                0.12           3.57          27.08
             FEM64                4 096   1.0915E-012                0.93           7.48           8.06
             FEM128              16 384   1.5655E-012                6.08           6.50           5.44
             FEM256              65 536   1.1976E-011               52.40           8.62           5.22


        d = 4, c = 0.2, eig = 10−4 , N(ni ) = ni (log2 ni ) Csp Cid
        Time t only H-PINVIT (without H-inversion)


Max Planck Institute Magdeburg            Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   25/28
Hierarchical (H-)Matrices                            PINVIT                                        Numerical Results



   2D Laplace, inner eigenvalues
        2D Laplace over [−1, 1] × [−1, 1], precond.: H-Cholesky decomp.

                                                                                   ti          N(ni )
             Name                 ni       error                     ti          ti−1         N(ni−1 )
             FEM8                 64   1.5306E-013              <0.01
             FEM16               256   9.8071E-012                0.02                        106.67
             FEM32             1 024   1.0577E-012                0.05           2.50          27.08
             FEM64             4 096   1.5776E-012                0.31           6.20           8.06
             FEM128           16 384   2.1416E-012                1.64           5.29           5.44
             FEM256           65 536   6.7400E-012                9.73           5.93           5.22
             FEM512          262 144   1.5002E-010              545.00          56.01           6.51

        d = 4, c = 0.2, eig = 10−4 , N(ni ) = ni (log2 ni ) Csp Cid
        Time t only H-PINVIT (without H-Cholesky decomposition)


Max Planck Institute Magdeburg         Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices   25/28
Preconditioned Inverse Iteration for Hierarchical Matrices
Preconditioned Inverse Iteration for Hierarchical Matrices
Preconditioned Inverse Iteration for Hierarchical Matrices
Preconditioned Inverse Iteration for Hierarchical Matrices
Preconditioned Inverse Iteration for Hierarchical Matrices
Preconditioned Inverse Iteration for Hierarchical Matrices
Preconditioned Inverse Iteration for Hierarchical Matrices
Preconditioned Inverse Iteration for Hierarchical Matrices
Preconditioned Inverse Iteration for Hierarchical Matrices
Preconditioned Inverse Iteration for Hierarchical Matrices
Preconditioned Inverse Iteration for Hierarchical Matrices
Preconditioned Inverse Iteration for Hierarchical Matrices
Preconditioned Inverse Iteration for Hierarchical Matrices
Preconditioned Inverse Iteration for Hierarchical Matrices

Mais conteúdo relacionado

Semelhante a Preconditioned Inverse Iteration for Hierarchical Matrices

Eigenvalues of Symmetrix Hierarchical Matrices
Eigenvalues of Symmetrix Hierarchical MatricesEigenvalues of Symmetrix Hierarchical Matrices
Eigenvalues of Symmetrix Hierarchical MatricesThomas Mach
 
Overview of sparse and low-rank matrix / tensor techniques
Overview of sparse and low-rank matrix / tensor techniques Overview of sparse and low-rank matrix / tensor techniques
Overview of sparse and low-rank matrix / tensor techniques Alexander Litvinenko
 
Vision systems_Image processing tool box in MATLAB
Vision systems_Image processing tool box in MATLABVision systems_Image processing tool box in MATLAB
Vision systems_Image processing tool box in MATLABHinna Nayab
 
IJERD(www.ijerd.com)International Journal of Engineering Research and Develop...
IJERD(www.ijerd.com)International Journal of Engineering Research and Develop...IJERD(www.ijerd.com)International Journal of Engineering Research and Develop...
IJERD(www.ijerd.com)International Journal of Engineering Research and Develop...IJERD Editor
 
IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...
IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...
IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...IJERD Editor
 
Skiena algorithm 2007 lecture17 edit distance
Skiena algorithm 2007 lecture17 edit distanceSkiena algorithm 2007 lecture17 edit distance
Skiena algorithm 2007 lecture17 edit distancezukun
 
Radial basis function neural network control for parallel spatial robot
Radial basis function neural network control for parallel spatial robotRadial basis function neural network control for parallel spatial robot
Radial basis function neural network control for parallel spatial robotTELKOMNIKA JOURNAL
 

Semelhante a Preconditioned Inverse Iteration for Hierarchical Matrices (9)

Eigenvalues of Symmetrix Hierarchical Matrices
Eigenvalues of Symmetrix Hierarchical MatricesEigenvalues of Symmetrix Hierarchical Matrices
Eigenvalues of Symmetrix Hierarchical Matrices
 
Matlab matrics
Matlab matricsMatlab matrics
Matlab matrics
 
Overview of sparse and low-rank matrix / tensor techniques
Overview of sparse and low-rank matrix / tensor techniques Overview of sparse and low-rank matrix / tensor techniques
Overview of sparse and low-rank matrix / tensor techniques
 
Vision systems_Image processing tool box in MATLAB
Vision systems_Image processing tool box in MATLABVision systems_Image processing tool box in MATLAB
Vision systems_Image processing tool box in MATLAB
 
IJERD(www.ijerd.com)International Journal of Engineering Research and Develop...
IJERD(www.ijerd.com)International Journal of Engineering Research and Develop...IJERD(www.ijerd.com)International Journal of Engineering Research and Develop...
IJERD(www.ijerd.com)International Journal of Engineering Research and Develop...
 
IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...
IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...
IJERD (www.ijerd.com) International Journal of Engineering Research and Devel...
 
Skiena algorithm 2007 lecture17 edit distance
Skiena algorithm 2007 lecture17 edit distanceSkiena algorithm 2007 lecture17 edit distance
Skiena algorithm 2007 lecture17 edit distance
 
Radial basis function neural network control for parallel spatial robot
Radial basis function neural network control for parallel spatial robotRadial basis function neural network control for parallel spatial robot
Radial basis function neural network control for parallel spatial robot
 
Clustering-beamer.pdf
Clustering-beamer.pdfClustering-beamer.pdf
Clustering-beamer.pdf
 

Mais de Thomas Mach

Fast and backward stable computation of roots of polynomials
Fast and backward stable computation of roots of polynomialsFast and backward stable computation of roots of polynomials
Fast and backward stable computation of roots of polynomialsThomas Mach
 
On Deflations in Extended QR Algorithms
On Deflations in Extended QR AlgorithmsOn Deflations in Extended QR Algorithms
On Deflations in Extended QR AlgorithmsThomas Mach
 
On Deflations in Extended QR Algorithms
On Deflations in Extended QR AlgorithmsOn Deflations in Extended QR Algorithms
On Deflations in Extended QR AlgorithmsThomas Mach
 
ADI for Tensor Structured Equations
ADI for Tensor Structured EquationsADI for Tensor Structured Equations
ADI for Tensor Structured EquationsThomas Mach
 
ADI for Tensor Structured Equations
ADI for Tensor Structured Equations ADI for Tensor Structured Equations
ADI for Tensor Structured Equations Thomas Mach
 
Computing Inner Eigenvalues of Matrices in Tensor Train Matrix Format
Computing Inner Eigenvalues of Matrices in Tensor Train Matrix FormatComputing Inner Eigenvalues of Matrices in Tensor Train Matrix Format
Computing Inner Eigenvalues of Matrices in Tensor Train Matrix FormatThomas Mach
 
Hierarchical Matrices: Concept, Application and Eigenvalues
Hierarchical Matrices: Concept, Application and EigenvaluesHierarchical Matrices: Concept, Application and Eigenvalues
Hierarchical Matrices: Concept, Application and EigenvaluesThomas Mach
 

Mais de Thomas Mach (7)

Fast and backward stable computation of roots of polynomials
Fast and backward stable computation of roots of polynomialsFast and backward stable computation of roots of polynomials
Fast and backward stable computation of roots of polynomials
 
On Deflations in Extended QR Algorithms
On Deflations in Extended QR AlgorithmsOn Deflations in Extended QR Algorithms
On Deflations in Extended QR Algorithms
 
On Deflations in Extended QR Algorithms
On Deflations in Extended QR AlgorithmsOn Deflations in Extended QR Algorithms
On Deflations in Extended QR Algorithms
 
ADI for Tensor Structured Equations
ADI for Tensor Structured EquationsADI for Tensor Structured Equations
ADI for Tensor Structured Equations
 
ADI for Tensor Structured Equations
ADI for Tensor Structured Equations ADI for Tensor Structured Equations
ADI for Tensor Structured Equations
 
Computing Inner Eigenvalues of Matrices in Tensor Train Matrix Format
Computing Inner Eigenvalues of Matrices in Tensor Train Matrix FormatComputing Inner Eigenvalues of Matrices in Tensor Train Matrix Format
Computing Inner Eigenvalues of Matrices in Tensor Train Matrix Format
 
Hierarchical Matrices: Concept, Application and Eigenvalues
Hierarchical Matrices: Concept, Application and EigenvaluesHierarchical Matrices: Concept, Application and Eigenvalues
Hierarchical Matrices: Concept, Application and Eigenvalues
 

Último

Web & Social Media Analytics Previous Year Question Paper.pdf
Web & Social Media Analytics Previous Year Question Paper.pdfWeb & Social Media Analytics Previous Year Question Paper.pdf
Web & Social Media Analytics Previous Year Question Paper.pdfJayanti Pande
 
The basics of sentences session 2pptx copy.pptx
The basics of sentences session 2pptx copy.pptxThe basics of sentences session 2pptx copy.pptx
The basics of sentences session 2pptx copy.pptxheathfieldcps1
 
Sports & Fitness Value Added Course FY..
Sports & Fitness Value Added Course FY..Sports & Fitness Value Added Course FY..
Sports & Fitness Value Added Course FY..Disha Kariya
 
APM Welcome, APM North West Network Conference, Synergies Across Sectors
APM Welcome, APM North West Network Conference, Synergies Across SectorsAPM Welcome, APM North West Network Conference, Synergies Across Sectors
APM Welcome, APM North West Network Conference, Synergies Across SectorsAssociation for Project Management
 
mini mental status format.docx
mini    mental       status     format.docxmini    mental       status     format.docx
mini mental status format.docxPoojaSen20
 
A Critique of the Proposed National Education Policy Reform
A Critique of the Proposed National Education Policy ReformA Critique of the Proposed National Education Policy Reform
A Critique of the Proposed National Education Policy ReformChameera Dedduwage
 
1029-Danh muc Sach Giao Khoa khoi 6.pdf
1029-Danh muc Sach Giao Khoa khoi  6.pdf1029-Danh muc Sach Giao Khoa khoi  6.pdf
1029-Danh muc Sach Giao Khoa khoi 6.pdfQucHHunhnh
 
Q4-W6-Restating Informational Text Grade 3
Q4-W6-Restating Informational Text Grade 3Q4-W6-Restating Informational Text Grade 3
Q4-W6-Restating Informational Text Grade 3JemimahLaneBuaron
 
Organic Name Reactions for the students and aspirants of Chemistry12th.pptx
Organic Name Reactions  for the students and aspirants of Chemistry12th.pptxOrganic Name Reactions  for the students and aspirants of Chemistry12th.pptx
Organic Name Reactions for the students and aspirants of Chemistry12th.pptxVS Mahajan Coaching Centre
 
BAG TECHNIQUE Bag technique-a tool making use of public health bag through wh...
BAG TECHNIQUE Bag technique-a tool making use of public health bag through wh...BAG TECHNIQUE Bag technique-a tool making use of public health bag through wh...
BAG TECHNIQUE Bag technique-a tool making use of public health bag through wh...Sapna Thakur
 
BASLIQ CURRENT LOOKBOOK LOOKBOOK(1) (1).pdf
BASLIQ CURRENT LOOKBOOK  LOOKBOOK(1) (1).pdfBASLIQ CURRENT LOOKBOOK  LOOKBOOK(1) (1).pdf
BASLIQ CURRENT LOOKBOOK LOOKBOOK(1) (1).pdfSoniaTolstoy
 
Sanyam Choudhary Chemistry practical.pdf
Sanyam Choudhary Chemistry practical.pdfSanyam Choudhary Chemistry practical.pdf
Sanyam Choudhary Chemistry practical.pdfsanyamsingh5019
 
The Most Excellent Way | 1 Corinthians 13
The Most Excellent Way | 1 Corinthians 13The Most Excellent Way | 1 Corinthians 13
The Most Excellent Way | 1 Corinthians 13Steve Thomason
 
Mastering the Unannounced Regulatory Inspection
Mastering the Unannounced Regulatory InspectionMastering the Unannounced Regulatory Inspection
Mastering the Unannounced Regulatory InspectionSafetyChain Software
 
Paris 2024 Olympic Geographies - an activity
Paris 2024 Olympic Geographies - an activityParis 2024 Olympic Geographies - an activity
Paris 2024 Olympic Geographies - an activityGeoBlogs
 
Disha NEET Physics Guide for classes 11 and 12.pdf
Disha NEET Physics Guide for classes 11 and 12.pdfDisha NEET Physics Guide for classes 11 and 12.pdf
Disha NEET Physics Guide for classes 11 and 12.pdfchloefrazer622
 
Ecosystem Interactions Class Discussion Presentation in Blue Green Lined Styl...
Ecosystem Interactions Class Discussion Presentation in Blue Green Lined Styl...Ecosystem Interactions Class Discussion Presentation in Blue Green Lined Styl...
Ecosystem Interactions Class Discussion Presentation in Blue Green Lined Styl...fonyou31
 

Último (20)

Web & Social Media Analytics Previous Year Question Paper.pdf
Web & Social Media Analytics Previous Year Question Paper.pdfWeb & Social Media Analytics Previous Year Question Paper.pdf
Web & Social Media Analytics Previous Year Question Paper.pdf
 
The basics of sentences session 2pptx copy.pptx
The basics of sentences session 2pptx copy.pptxThe basics of sentences session 2pptx copy.pptx
The basics of sentences session 2pptx copy.pptx
 
Mattingly "AI & Prompt Design: Structured Data, Assistants, & RAG"
Mattingly "AI & Prompt Design: Structured Data, Assistants, & RAG"Mattingly "AI & Prompt Design: Structured Data, Assistants, & RAG"
Mattingly "AI & Prompt Design: Structured Data, Assistants, & RAG"
 
Sports & Fitness Value Added Course FY..
Sports & Fitness Value Added Course FY..Sports & Fitness Value Added Course FY..
Sports & Fitness Value Added Course FY..
 
APM Welcome, APM North West Network Conference, Synergies Across Sectors
APM Welcome, APM North West Network Conference, Synergies Across SectorsAPM Welcome, APM North West Network Conference, Synergies Across Sectors
APM Welcome, APM North West Network Conference, Synergies Across Sectors
 
mini mental status format.docx
mini    mental       status     format.docxmini    mental       status     format.docx
mini mental status format.docx
 
A Critique of the Proposed National Education Policy Reform
A Critique of the Proposed National Education Policy ReformA Critique of the Proposed National Education Policy Reform
A Critique of the Proposed National Education Policy Reform
 
Código Creativo y Arte de Software | Unidad 1
Código Creativo y Arte de Software | Unidad 1Código Creativo y Arte de Software | Unidad 1
Código Creativo y Arte de Software | Unidad 1
 
1029-Danh muc Sach Giao Khoa khoi 6.pdf
1029-Danh muc Sach Giao Khoa khoi  6.pdf1029-Danh muc Sach Giao Khoa khoi  6.pdf
1029-Danh muc Sach Giao Khoa khoi 6.pdf
 
Q4-W6-Restating Informational Text Grade 3
Q4-W6-Restating Informational Text Grade 3Q4-W6-Restating Informational Text Grade 3
Q4-W6-Restating Informational Text Grade 3
 
Organic Name Reactions for the students and aspirants of Chemistry12th.pptx
Organic Name Reactions  for the students and aspirants of Chemistry12th.pptxOrganic Name Reactions  for the students and aspirants of Chemistry12th.pptx
Organic Name Reactions for the students and aspirants of Chemistry12th.pptx
 
BAG TECHNIQUE Bag technique-a tool making use of public health bag through wh...
BAG TECHNIQUE Bag technique-a tool making use of public health bag through wh...BAG TECHNIQUE Bag technique-a tool making use of public health bag through wh...
BAG TECHNIQUE Bag technique-a tool making use of public health bag through wh...
 
BASLIQ CURRENT LOOKBOOK LOOKBOOK(1) (1).pdf
BASLIQ CURRENT LOOKBOOK  LOOKBOOK(1) (1).pdfBASLIQ CURRENT LOOKBOOK  LOOKBOOK(1) (1).pdf
BASLIQ CURRENT LOOKBOOK LOOKBOOK(1) (1).pdf
 
Sanyam Choudhary Chemistry practical.pdf
Sanyam Choudhary Chemistry practical.pdfSanyam Choudhary Chemistry practical.pdf
Sanyam Choudhary Chemistry practical.pdf
 
The Most Excellent Way | 1 Corinthians 13
The Most Excellent Way | 1 Corinthians 13The Most Excellent Way | 1 Corinthians 13
The Most Excellent Way | 1 Corinthians 13
 
Mastering the Unannounced Regulatory Inspection
Mastering the Unannounced Regulatory InspectionMastering the Unannounced Regulatory Inspection
Mastering the Unannounced Regulatory Inspection
 
Paris 2024 Olympic Geographies - an activity
Paris 2024 Olympic Geographies - an activityParis 2024 Olympic Geographies - an activity
Paris 2024 Olympic Geographies - an activity
 
Disha NEET Physics Guide for classes 11 and 12.pdf
Disha NEET Physics Guide for classes 11 and 12.pdfDisha NEET Physics Guide for classes 11 and 12.pdf
Disha NEET Physics Guide for classes 11 and 12.pdf
 
Ecosystem Interactions Class Discussion Presentation in Blue Green Lined Styl...
Ecosystem Interactions Class Discussion Presentation in Blue Green Lined Styl...Ecosystem Interactions Class Discussion Presentation in Blue Green Lined Styl...
Ecosystem Interactions Class Discussion Presentation in Blue Green Lined Styl...
 
INDIA QUIZ 2024 RLAC DELHI UNIVERSITY.pptx
INDIA QUIZ 2024 RLAC DELHI UNIVERSITY.pptxINDIA QUIZ 2024 RLAC DELHI UNIVERSITY.pptx
INDIA QUIZ 2024 RLAC DELHI UNIVERSITY.pptx
 

Preconditioned Inverse Iteration for Hierarchical Matrices

  • 1. 82nd GAMM Annual Scientific Conference Graz, April 19, 2011 Preconditioned Inverse Iteration for Hierarchical Matrices Peter Benner and Thomas Mach Max Planck Institute for Dynamics of Complex Technical Systems Computational Methods in Systems and Control Theory Magdeburg MAX PLANCK INSTITUTE FOR DYNAMICS OF COMPLEX TECHNICAL SYSTEMS MAGDEBURG Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 1/28
  • 2. Hierarchical (H-)Matrices PINVIT Numerical Results Outline 1 Hierarchical (H-)Matrices 2 PINVIT 3 Numerical Results Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 2/28
  • 3. Hierarchical (H-)Matrices PINVIT Numerical Results H-Matrices [Hackbusch 1998] Some dense matrices, e.g. BEM or FEM, can be approximated by H-matrices in a data-sparse manner. hierarchical tree TI block H-tree TI × I I = {1, 2, 3, 4, 5, 6, 7, 8} 12345678 12345678 12345678 12345678 1 1 1 1 2 2 2 2 {1, 2, 3, 4} {5, 6, 7, 8} 3 3 3 3 4 4 4 4 5 5 5 5 6 6 6 6 {1, 2} {3, 4} {5, 6} {7, 8} 7 7 7 7 8 8 8 8 {1}{2}{3}{4}{5}{6}{7}{8} dense matrices, rank-k-matrices rank-k-matrix: Ms×t = AB T , A ∈ Rn×k , B ∈ Rm×k (k n, m) Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 3/28
  • 4. Hierarchical (H-)Matrices PINVIT Numerical Results H-Matrices [Hackbusch 1998] Hierarchical matrices H(TI × I , k) = M ∈ RI × I rank (Ma×b ) ≤ k ∀a × b admissible 3 3 22 7 10 14 8 11 9 8 5 3 7 19 10 11 8 3 10 10 31 11 11 9 16 12 8 15 8 12 11 8 19 10 11 8 14 8 11 11 11 8 10 11 31 11 31 11 9 3 16 11 7 3 8 6 15 5 6 7 13 5 8 8 15 9 15 12 13 adaptive rank k(ε) 19 13 11 61 11 10 7 8 9 16 11 11 8 11 8 13 13 9 3 25 10 13 8 7 11 10 19 11 6 5 8 11 9 11 16 3 storage NSt,H (T , k) = O(n log n k(ε)) 8 11 11 31 11 15 6 9 15 11 11 8 8 8 15 10 10 15 9 5 8 11 61 3 6 14 9 11 10 10 7 13 6 15 10 3 6 25 10 6 13 8 12 5 6 10 14 6 10 19 10 10 31 10 16 9 11 8 8 15 11 12 20 13 8 complexity of approximate arithmetic 7 8 11 15 9 10 10 5 8 11 8 11 10 51 7 9 3 7 3 9 7 6 15 10 25 11 8 13 13 7 13 11 5 6 10 16 10 9 7 3 3 11 25 10 10 19 9 13 8 8 11 11 11 8 3 7 8 15 8 39 10 10 10 6 5 10 3 9 15 3 10 3 25 7 10 3 6 3 15 10 13 6 11 13 7 10 7 22 7 10 11 9 8 5 12 O(n log n k(ε)) 10 7 19 10 11 8 MH v 3 7 12 10 6 3 10 10 31 11 9 16 12 8 15 8 20 13 8 8 11 8 34 10 13 10 6 5 10 9 15 15 11 10 25 7 11 13 6 11 13 13 11 8 13 7 O(n log n k(ε)2 ) 7 11 10 16 61 11 23 +H , −H 9 10 11 13 9 6 13 20 9 9 7 11 8 3 7 9 8 8 15 5 6 12 9 39 10 3 10 15 10 12 13 11 8 8 8 15 9 10 15 9 7 11 11 3 3 61 10 −1 ∗H , HLU(·), (·)H O(n (log n)2 k(ε)2 ) 5 8 7 7 10 9 11 19 13 9 6 13 20 9 9 7 9 8 13 8 8 11 5 6 10 15 10 9 34 10 13 12 7 11 12 11 23 10 15 8 9 11 9 7 10 13 51 Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 4/28
  • 5. Hierarchical (H-)Matrices PINVIT Numerical Results H-Matrices [Hackbusch 1998] Hierarchical matrices H(TI × I , k) = M ∈ RI × I rank (Ma×b ) ≤ k ∀a × b admissible 3 3 22 7 10 14 8 11 9 8 5 3 7 19 10 11 8 3 10 10 31 11 11 9 16 12 8 15 8 12 11 8 19 10 11 8 14 8 11 11 11 8 10 11 31 11 31 11 9 3 16 11 7 3 8 6 15 5 6 7 13 5 8 8 15 9 15 12 13 adaptive rank k(ε) 19 13 11 61 11 10 7 8 9 16 11 11 8 11 8 13 13 9 3 25 10 13 8 7 11 10 19 11 6 5 8 11 9 11 16 3 storage NSt,H (T , k) = O(n log n k(ε)) 8 11 11 31 11 15 6 9 15 11 11 8 8 8 15 10 10 15 9 5 8 11 61 3 6 14 9 11 10 10 7 13 6 15 10 3 6 25 10 6 13 8 12 5 6 10 14 6 10 19 10 10 31 10 16 9 11 8 8 15 11 12 20 13 8 complexity of approximate arithmetic 7 8 11 15 9 10 10 5 8 11 8 11 10 51 7 9 3 7 3 9 7 6 15 10 25 11 8 13 13 7 13 11 5 6 10 16 10 9 7 3 3 11 25 10 10 19 9 13 8 8 11 11 11 8 3 7 8 15 8 39 10 10 10 6 5 10 3 9 15 3 10 3 25 7 10 3 6 3 15 10 13 6 11 13 7 10 7 22 7 10 11 9 8 5 12 O(n log n k(ε)) 10 7 19 10 11 8 MH v 3 7 12 10 6 3 10 10 31 11 9 16 12 8 15 8 20 13 8 8 11 8 34 10 13 10 6 5 10 9 15 15 11 10 25 7 11 13 6 11 13 13 11 8 13 7 O(n log n k(ε)2 ) 7 11 10 16 61 11 23 +H , −H 9 10 11 13 9 6 13 20 9 9 7 11 8 3 7 9 8 8 15 5 6 12 9 39 10 3 10 15 10 12 13 11 8 8 8 15 9 10 15 9 7 11 11 3 3 61 10 −1 ∗H , HLU(·), (·)H O(n (log n)2 k(ε)2 ) 5 8 7 7 10 9 11 19 13 9 6 13 20 9 9 7 9 8 13 8 8 11 5 6 10 15 10 9 34 10 13 12 7 11 12 11 23 10 15 8 9 11 9 7 10 13 51 T B1 T B2 T B1 A1 + A2 = A1 A2 T B2 Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 4/28
  • 6. Hierarchical (H-)Matrices PINVIT Numerical Results Goal: Eigenvalues of symmetric H-Matrices M = M T ∈ H(T , k), Λ(M) = {λ1, λ2, . . . , λn }, λ1 ≥ λ2 ≥ · · · ≥ λn > 0 ⇓ How to find λi in O(n (log n)α k β )? Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 5/28
  • 7. Hierarchical (H-)Matrices PINVIT Numerical Results Goal: Eigenvalues of symmetric H-Matrices M = M T ∈ H(T , k), Λ(M) = {λ1, λ2, . . . , λn }, λ1 ≥ λ2 ≥ · · · ≥ λn > 0 ⇓ How to find λn in O(n (log n)α k β )? Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 5/28
  • 8. Hierarchical (H-)Matrices PINVIT Numerical Results Preconditioned Inverse Iteration [Knyazev, Neymeyr, et al.] Definition The function x T Mx µ(x) = µ(x, M) = xT x is called the Rayleigh quotient. Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 6/28
  • 9. Hierarchical (H-)Matrices PINVIT Numerical Results Preconditioned Inverse Iteration [Knyazev, Neymeyr, et al.] Definition The function x T Mx µ(x) = µ(x, M) = xT x is called the Rayleigh quotient. Minimize the Rayleigh quotient by a gradient method: 2 xi+1 := xi − α µ(xi ), µ(x) = T (Mx − xµ(x)) , x x Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 6/28
  • 10. Hierarchical (H-)Matrices PINVIT Numerical Results Preconditioned Inverse Iteration [Knyazev, Neymeyr, et al.] Residual r (x) = Mx − xµ(x). Minimize the Rayleigh quotient by a gradient method: 2 xi+1 := xi − α µ(xi ), µ(x) = T (Mx − xµ(x)) , x x Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 6/28
  • 11. Hierarchical (H-)Matrices PINVIT Numerical Results Preconditioned Inverse Iteration [Knyazev, Neymeyr, et al.] Definition The function x T Mx µ(x) = µ(x, M) = xT x is called the Rayleigh quotient. Minimize the Rayleigh quotient by a gradient method: 2 xi+1 := xi − α µ(xi ), µ(x) = T (Mx − xµ(x)) , x x + preconditioning ⇒ update equation: xi+1 := xi − B −1 (Mxi − xi µ(xi )) . Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 6/28
  • 12. Hierarchical (H-)Matrices PINVIT Numerical Results Preconditioned Inverse Iteration [Knyazev, Neymeyr, et al.] Preconditioned residual B −1 r (x) = B −1 (Mx − xµ(x)). + preconditioning ⇒ update equation: xi+1 := xi − B −1 (Mxi − xi µ(xi )) . Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 6/28
  • 13. Hierarchical (H-)Matrices PINVIT Numerical Results Preconditioned Inverse Iteration [Knyazev, Neymeyr 2009] xi+1 := xi − B −1 (Mxi − xi µ(xi )) If M ∈ Rn×n symmetric positive definite and B −1 approximates the inverse of M, so that I − B −1 M M ≤ c < 1, then Preconditioned INVerse ITeration (PINVIT) converges and the number of iterations is independent of n. Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 7/28
  • 14. Hierarchical (H-)Matrices PINVIT Numerical Results Preconditioned Inverse Iteration [Knyazev, Neymeyr, et al.] The residual ri = Mxi − xi µ(xi ) converges to 0, so that ri 2 < is a useful termination criterion. Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 8/28
  • 15. Hierarchical (H-)Matrices PINVIT Numerical Results Variants of PINVIT [Neymeyr 2001: A Hierarchy of Precond. Eigens. for Ellipt. Diff. Op.] Classification by Neymeyr: PINVIT(1): xi+1 := xi − B −1 ri . PINVIT(2): xi+1 := arg minv ∈span{xi ,B −1 ri } µ(v ). PINVIT(3): xi+1 := arg minv ∈span{xi−1 ,xi ,B −1 ri } µ(v ). PINVIT(n): Analogously. PINVIT(·,d): Replacing x by a rectangular full rank matrix X ∈ Rn×d one gets the subspace version of PINVIT(·). Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 9/28
  • 16. Hierarchical (H-)Matrices PINVIT Numerical Results Algorithm and Complexity The number of iterations is independent of matrix size n. H-PINVIT(1,d) Input: M ∈ Rn×n , X0 ∈ Rn×d (X0 X0 = I , e.g. randomly chosen) T Output: Xp ∈ R n×d , µ ∈ Rd×d , with MX − X µ ≤ p p B −1 = (M)−1 or B −1 = L−T L−1 H H H T R := MX0 − X0 µ, µ = X0 MX0 for (i := 1; R F > ; i + +) do Xi := Orthogonalize Xi−1 − B −1 R R := MXi − Xi µ, µ = XiT MXi end Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 10/28
  • 17. Hierarchical (H-)Matrices PINVIT Numerical Results Algorithm and Complexity The number of iterations is independent of matrix size n. H-PINVIT(1,d) Input: M ∈ Rn×n , X0 ∈ Rn×d (X0 X0 = I , e.g. randomly chosen) T Output: Xp ∈ R n×d , µ ∈ Rd×d , with MX − X µ ≤ p p B −1 = (M)−1 or B −1 = L−T L−1 H H H O(n (log n)2 k (c)2 ) T R := MX0 − X0 µ, µ = X0 MX0 for (i := 1; R F > ; i + +) do Xi := Orthogonalize Xi−1 − B −1 R O(n (log n) k (c)2 ) R := MXi − Xi µ, µ = XiT MXi O(n (log n) k (c)) end The complexity of the algorithm is determined by the H-matrix inversion/Cholesky decomposition: ⇒ O(n (log n)2 k (c)2 ). Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 10/28
  • 18. Hierarchical (H-)Matrices PINVIT Numerical Results Goal: Eigenvalues of symmetric H-Matrices M = M T ∈ H(T , k), Λ(M) = {λ1, λ2, . . . , λn }, λ1 ≥ λ2 ≥ · · · ≥ λn > 0 ⇓ How to find λn in O(n (log n)α k β )? Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 11/28
  • 19. Hierarchical (H-)Matrices PINVIT Numerical Results Goal: Eigenvalues of symmetric H-Matrices M = M T ∈ H(T , k), Λ(M) = {λ1, λ2, . . . , λn }, λ1 ≥ λ2 ≥ · · · ≥ λn > 0 ⇓ How to find λn in O(n (log n)α k β )? Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 11/28
  • 20. Hierarchical (H-)Matrices PINVIT Numerical Results Goal: Eigenvalues of symmetric H-Matrices M = M T ∈ H(T , k), Λ(M) = {λ1, λ2, . . . , λn }, λ1 ≥ λ2 ≥ · · · ≥ λn > 0 ⇓ How to find λn in O(n (log n)α k β )? How to find λi in O(n (log n)α k β )? Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 12/28
  • 21. Hierarchical (H-)Matrices PINVIT Numerical Results Folded Spectrum Method How to find λi in O(n (log n)α k β )? If i = n − d with d < O(log n), use subspace version PINVIT(·,d ). ... 0λn λn−1 λn−2 λ1 Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 13/28
  • 22. Hierarchical (H-)Matrices PINVIT Numerical Results Folded Spectrum Method How to find λi in O(n (log n)α k β )? If i = n − d with d log n? ... ... 0λn λi+1 λi λi−1 λ1 Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 13/28
  • 23. Hierarchical (H-)Matrices PINVIT Numerical Results Folded Spectrum Method How to find λi in O(n (log n)α k β )? If i = n − d with d log n, shift with σ near λi . ... ... 0λn λi+1 λi λi−1 λ1 σ Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 13/28
  • 24. Hierarchical (H-)Matrices PINVIT Numerical Results Folded Spectrum Method How to find λi in O(n (log n)α k β )? If i = n − d with d log n, shift with σ near λi . ... ... 0λn λi+1 λi λi−1 λ1 σ But (M − σI) is not positive definite. Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 13/28
  • 25. Hierarchical (H-)Matrices PINVIT Numerical Results Folded Spectrum Method Folded Spectrum Method [Wang, Zunger 1994] Mσ = (M − σI)2 Mσ is s.p.d., if M is s.p.d. and σ = λi . Assume all eigenvalues of Mσ are simple. Mv = λv ⇔ Mσ v = (M − σI)2 v = M 2 v − 2σMv + σ 2 v = λ2 v − 2σλv + σ 2 v = (λ − σ)2 v Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 14/28
  • 26. Hierarchical (H-)Matrices PINVIT Numerical Results Folded Spectrum Method 1 Choose σ. 2 Compute 2 a) Mσ = (M − σI) and −1 b) Mσ or LLT = Mσ . 3 Use PINVIT to find the smallest eigenpair (µσ , v ) of Mσ . 4 Compute µ = v T Mv /v T v . (µ, v ) is the nearest eigenpair to σ. Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 15/28
  • 27. Hierarchical (H-)Matrices PINVIT Numerical Results Folded Spectrum Method 1 Choose σ. 2 Compute 2 a) Mσ = (M − σI) and −1 b) Mσ or LLT = Mσ . 3 Use PINVIT to find the smallest eigenpair (µσ , v ) of Mσ . 4 Compute µ = v T Mv /v T v . (µ, v ) is the nearest eigenpair to σ. H-arithmetic enables us to shift, square and invert M resp. Mσ in linear-polylogarithmic complexity. Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 15/28
  • 28. Hierarchical (H-)Matrices PINVIT Numerical Results Folded Spectrum Method 1 Choose σ. 2 Compute 2 a) Mσ = (M − σI) and −1 b) Mσ or LLT = Mσ . 3 Use PINVIT to find the smallest eigenpair (µσ , v ) of Mσ . 4 Compute µ = v T Mv /v T v . (µ, v ) is the nearest eigenpair to σ. H-arithmetic enables us to shift, square and invert M resp. Mσ in linear-polylogarithmic complexity. If M is sparse, then shifting, squaring and inverting is prohibitive. Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 15/28
  • 29. Hierarchical (H-)Matrices PINVIT Numerical Results Folded Spectrum Method 1 Choose σ. 2 Compute 2 a) Mσ = (M − σI) and −1 b) Mσ or LLT = Mσ . 3 Use PINVIT to find the smallest eigenpair (µσ , v ) of Mσ . 4 Compute µ = v T Mv /v T v . (µ, v ) is the nearest eigenpair to σ. H-arithmetic enables us to shift, square and invert M resp. Mσ in linear-polylogarithmic complexity. If M is sparse, then shifting, squaring and inverting is prohibitive. Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 15/28
  • 30. Hierarchical (H-)Matrices PINVIT Numerical Results Folded Spectrum Method 1 Choose σ. 2 Compute 2 a) Mσ = (M − σI) and −1 b) Mσ or LLT = Mσ . 3 Use PINVIT to find the smallest eigenpair (µσ , v ) of Mσ . 4 Compute µ = v T Mv /v T v . (µ, v ) is the nearest eigenpair to σ. H-arithmetic enables us to shift, square and invert M resp. Mσ in linear-polylogarithmic complexity. If M is sparse, then shifting, squaring and inverting is prohibitive. Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 15/28
  • 31. Hierarchical (H-)Matrices PINVIT Numerical Results Goal: Eigenvalues of symmetric H-Matrices M = M T ∈ H(T , k), Λ(M) = {λ1, λ2, . . . , λn }, λ1 ≥ λ2 ≥ · · · ≥ λn > 0 ⇓ How to find λn in O(n (log n)α k β )? How to find λi in O(n (log n)α k β )? Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 16/28
  • 32. Hierarchical (H-)Matrices PINVIT Numerical Results Goal: Eigenvalues of symmetric H-Matrices M = M T ∈ H(T , k), Λ(M) = {λ1, λ2, . . . , λn }, λ1 ≥ λ2 ≥ · · · ≥ λn > 0 ⇓ How to find λn in O(n (log n)α k β )? How to find λi in O(n (log n)α k β )? Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 16/28
  • 33. Hierarchical (H-)Matrices PINVIT Numerical Results Numerical Results Numerical Results Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 17/28
  • 34. Hierarchical (H-)Matrices PINVIT Numerical Results Hlib Hlib ¨ [Borm, Grasedyck, et al.] We use the Hlib1.3 (www.hlib.org) for the H-arithmetic operations and some examples out of the library for testing the eigenvalue algorithm. Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 18/28
  • 35. Hierarchical (H-)Matrices PINVIT Numerical Results 2D Laplace, smallest eigenvalues 2D Laplace over [−1, 1] × [−1, 1], precond.: H-inversion ti N(ni ) Name ni error ti ti−1 N(ni−1 ) FEM8 64 5.6146E-010 0.01 FEM16 256 4.5918E-010 0.02 2.00 106.67 FEM32 1 024 3.7550E-010 0.12 6.17 27.08 FEM64 4 096 3.8009E-010 0.82 6.68 8.06 FEM128 16 384 4.4099E-010 5.84 7.09 5.44 FEM256 65 536 3.9651E-010 34.47 5.91 5.22 FEM512 262 144 3.7877E-010 194.00 5.63 6.51 d = 4, c = 0.2, eig = 10−4 , N(ni ) = ni (log2 ni ) Csp Cid Time t only H-PINVIT (without H-inversion) Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 19/28
  • 36. Hierarchical (H-)Matrices PINVIT Numerical Results 2D Laplace, smallest eigenvalues 2D Laplace over [−1, 1] × [−1, 1], precond.: H-inversion ti N(ni ) Name ni error ti ti−1 N(ni−1 ) FEM8 64 5.6146E-010 0.01 FEM16 256 4.5918E-010 0.02 2.00 106.67 FEM32 1 024 3.7550E-010 0.12 6.17   27.08 FEM64 4 096 3.8009E-010 ˆ 0.821 − λ1 λ 6.68 8.06 FEM128 16 384 4.4099E-010  ˆ λ2 − λ2  5.84 7.09  5.44 FEM256 65 536 3.9651E-010 λ 5.91  ˆ 34.473 − λ3  5.22 FEM512 262 144 3.7877E-010 ˆ4 194.004 − λ λ 5.63 2 6.51 d = 4, c = 0.2, eig = 10−4 , N(ni ) = ni (log2 ni ) Csp Cid Time t only H-PINVIT (without H-inversion) Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 19/28
  • 37. Hierarchical (H-)Matrices PINVIT Numerical Results 2D Laplace, smallest eigenvalues 104 H-PINVIT(1,4) H-inversion 103 H-PINVIT(3,4) O(N(ni ) log ni ) O(N(ni )) 102 CPU time in s 101 100 10−1 10−2 10−3 16 32 64 128 EM256 EM512 FEM FEM FEM FEM F F Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 20/28
  • 38. Hierarchical (H-)Matrices PINVIT Numerical Results 2D Laplace, smallest eigenvalues 104 H-PINVIT(1,4) H-inversion 103 H-PINVIT(3,4) O(N(ni ) log ni ) O(N(ni )) MATLAB eigs 102 CPU time in s 101 100 10−1 10−2 10−3 16 32 64 128 EM256 EM512 FEM FEM FEM FEM F F Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 20/28
  • 39. Hierarchical (H-)Matrices PINVIT Numerical Results Algorithm and Complexity H-PINVIT(1,d) Input: M ∈ Rn×n , X0 ∈ Rn×d (X0 X0 = I , e.g. randomly chosen) T Output: Xp ∈ R n×d , µ ∈ Rd×d , with MX − X µ ≤ p p B −1 = (M)−1 or B −1 = L−T L−1 H H H O(n (log n)2 k (c)2 ) T R := MX0 − X0 µ, µ = X0 MX0 for (i := 1; R F > ; i + +) do Xi := Orthogonalize Xi−1 − B −1 R O(n (log n) k (c)2 ) R := MXi − Xi µ, µ = XiT MXi O(n (log n) k (c)) end Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 21/28
  • 40. Hierarchical (H-)Matrices PINVIT Numerical Results Algorithm and Complexity H-PINVIT(1,d) Input: MCompetitive to Rn×d (X0 X0 = I , e.g. randomly chosen) ∈ Rn×n , X0 ∈ MATLAB eigs. T Output: Xp ∈ R n×d , µ ∈ Rd×d , with MX − X µ ≤ p p B −1 = (M)−1 or B −1 = L−T L−1 H H H O(n (log n)2 k (c)2 ) T R := MX0 − X0 µ, µ = X0 MX0 for (i := 1; R F > ; i + +) do Xi := Orthogonalize Xi−1 − B −1 R O(n (log n) k (c)2 ) R := MXi − Xi µ, µ = XiT MXi O(n (log n) k (c)) end Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 21/28
  • 41. Hierarchical (H-)Matrices PINVIT Numerical Results Algorithm and Complexity H-PINVIT(1,d) Input: M ∈ Rn×n , X0 ∈ Rn×d (X0 X0 = I , e.g. randomly chosen) T Output: Xp ∈ R n×d , µ ∈ Rd×d , with MX − X µ ≤ p p B −1 = (M)−1 or B −1 = L−T L−1 H H H O(n (log n)2 k (c)2 ) T R := MX0 − X0 µ, µ = X0 MX0 for (i := 1; R F > ; i + +) do Xi := Orthogonalize Xi−1 − B −1 R O(n (log n) k (c)2 ) R := MXi − Xi µ, µ = XiT MXi O(n (log n) k (c)) end Expensive. Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 21/28
  • 42. Hierarchical (H-)Matrices PINVIT Numerical Results 2D Laplace, smallest eigenvalues 2D Laplace over [−1, 1] × [−1, 1], precond.: H-Cholesky decomp. ti N(ni ) Name ni error ti ti−1 N(ni−1 ) FEM8 64 4.6920E-010 0.01 FEM16 256 4.7963E-010 0.02 2.00 106.67 FEM32 1 024 3.4696E-010 0.08 4.00 27.08 FEM64 4 096 4.6414E-010 0.48 6.00 8.06 FEM128 16 384 3.3206E-010 3.20 6.67 5.44 FEM256 65 536 3.8468E-010 13.90 4.34 5.22 FEM512 262 144 3.1353E-010 62.40 4.49 6.51 d = 4, c = 0.2, eig = 10−4 , N(ni ) = ni (log2 ni ) Csp Cid Time t only H-PINVIT (without H-Cholesky decomposition) Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 22/28
  • 43. Hierarchical (H-)Matrices PINVIT Numerical Results 2D Laplace, smallest eigenvalues 104 H-PINVIT(1,4) H-Chol. decomp. 103 H-PINVIT(3,4) O(N(ni ) log ni ) O(N(ni )) 102 CPU time in s 101 100 10−1 10−2 10−3 16 32 64 128 EM256 EM512 FEM FEM FEM FEM F F Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 23/28
  • 44. Hierarchical (H-)Matrices PINVIT Numerical Results 2D Laplace, smallest eigenvalues 104 H-PINVIT(1,4) H-Chol. decomp. 103 H-PINVIT(3,4) O(N(ni ) log ni ) O(N(ni )) MATLAB eigs 102 CPU time in s 101 100 10−1 10−2 10−3 16 32 64 128 EM256 EM512 FEM FEM FEM FEM F F Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 23/28
  • 45. Hierarchical (H-)Matrices PINVIT Numerical Results 2D Laplace, vn , FEM64 ·10−2 6 4 ·10−2 5 2 0 0 −2 1 −5 −4 −1 0 −0.5 0 0.5 −6 1 −1 Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 24/28
  • 46. Hierarchical (H-)Matrices PINVIT Numerical Results 2D Laplace, vn−1 , FEM64 ·10−2 6 4 ·10−2 5 2 0 0 −2 1 −5 −4 −1 0 −0.5 0 0.5 −6 1 −1 Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 24/28
  • 47. Hierarchical (H-)Matrices PINVIT Numerical Results 2D Laplace, vn−2 , FEM64 ·10−2 6 4 ·10−2 5 2 0 0 −2 1 −5 −4 −1 0 −0.5 0 0.5 −6 1 −1 Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 24/28
  • 48. Hierarchical (H-)Matrices PINVIT Numerical Results 2D Laplace, vn−3 , FEM64 ·10−2 6 4 ·10−2 5 2 0 0 −2 1 −5 −4 −1 0 −0.5 0 0.5 −6 1 −1 Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 24/28
  • 49. Hierarchical (H-)Matrices PINVIT Numerical Results 2D Laplace, inner eigenvalues 2D Laplace over [−1, 1] × [−1, 1], precond.: H-inversion ti N(ni ) Name ni error ti ti−1 N(ni−1 ) FEM8 64 1.7219E-013 <0.01 FEM16 256 7.2388E-012 0.03 106.67 FEM32 1 024 6.0045E-013 0.12 3.57 27.08 FEM64 4 096 1.0915E-012 0.93 7.48 8.06 FEM128 16 384 1.5655E-012 6.08 6.50 5.44 FEM256 65 536 1.1976E-011 52.40 8.62 5.22 d = 4, c = 0.2, eig = 10−4 , N(ni ) = ni (log2 ni ) Csp Cid Time t only H-PINVIT (without H-inversion) Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 25/28
  • 50. Hierarchical (H-)Matrices PINVIT Numerical Results 2D Laplace, inner eigenvalues 2D Laplace over [−1, 1] × [−1, 1], precond.: H-Cholesky decomp. ti N(ni ) Name ni error ti ti−1 N(ni−1 ) FEM8 64 1.5306E-013 <0.01 FEM16 256 9.8071E-012 0.02 106.67 FEM32 1 024 1.0577E-012 0.05 2.50 27.08 FEM64 4 096 1.5776E-012 0.31 6.20 8.06 FEM128 16 384 2.1416E-012 1.64 5.29 5.44 FEM256 65 536 6.7400E-012 9.73 5.93 5.22 FEM512 262 144 1.5002E-010 545.00 56.01 6.51 d = 4, c = 0.2, eig = 10−4 , N(ni ) = ni (log2 ni ) Csp Cid Time t only H-PINVIT (without H-Cholesky decomposition) Max Planck Institute Magdeburg Thomas Mach, Preconditioned Inverse Iteration for Hierarchical Matrices 25/28