Group Nemeth
| Group Description |
A set of matrices from Karoly Nemeth, theoretical chemistry.
currently postdoctoral fellow at Rice University with Prof. G.E. Scuseria
(as of Summer 1999).
Symmetric-positive semidefinite matrices.
--------------------------------------------------------------------------------
Dear Prof Davis,
The problem, where my matrices come from, is the Newton-Schultz iteration
(NSI), which is used for the calculation of the matrix sign function.
If we denote the "sign" of a hermitian matrix "F" by "sign(F)",
sign(F) can be calculated by the NSI, like:
Z_{k+1} = Z_{k} + 1/2 * (I - Z_{k}^2) * Z_{k} .
"Z" is a general symmetric matrix, "k" is the iteration step, "I" is the
identity, Z_{0} = F/||F||, where ||F|| is the absolute value of the
largest eigenvalue of F. In my problems NSI converged in most cases in
less than 20 iterations.
Currently I am very much interested in the Cholesky-decomposition of
(I - Z_{k}^2)^2: (I - Z_{k}^2)^2 = T^{t}T (T is upper triangular) . This
would help me to construct some orthogonal vectors, which are important
for some applications, the vectors are generated by :
(I - Z_{k}^2)*T^{-1}.
As NSI converges, the eigenvalues in Z_{k} tend to +1 or -1. (I - Z_{k}^2)
gets more and more rank deficient (Z_{k}^2 tend to I). Thus normal Cholesky
cannot be used for the decomposition of (I - Z_{k}^2)^2. This is why I
need Cholesky factorization for positive semidefinite matrices. Thus I
expect to be able to generate "r" orthonormal vectors from (I - Z_{k}^2),
where "r" is the rank of (I - Z_{k}^2).
Here I used a somewhat
modified version of NSI, where the trace of the matrices is kept fixed.
However the matrices needed to be Cholesky decomposed are the corresponding
(1-Z_{k}^2)^2 matrices.
Best regards:
Karoly
(The Nemeth{K}.rsa file holds the Z_{k} matrix.)
***********************************
Karoly Nemeth, Ph.D.
----------------------------------
Rice University
Department of Chemistry - MS 60
6100 Main Street
Houston, TX 77005-1892
e-mail: karoly :at the domain: celaeno.rice.edu
office phone: before 5.30pm : +713-527-8101/2826
after 5.30pm : +713-527-8750/2826
fax: +713-285-5155
***********************************
|
|---|
Displaying collection matrices 1 - 20 of 26 in total
| Id | Name | Group | Rows | Cols | Nonzeros | Kind | Date | Download File |
|---|---|---|---|---|---|---|---|---|
| 780 | nemeth16 | Nemeth | 9,506 | 9,506 | 587,012 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 781 | nemeth17 | Nemeth | 9,506 | 9,506 | 629,620 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 782 | nemeth18 | Nemeth | 9,506 | 9,506 | 695,234 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 783 | nemeth19 | Nemeth | 9,506 | 9,506 | 818,302 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 784 | nemeth20 | Nemeth | 9,506 | 9,506 | 971,870 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 785 | nemeth21 | Nemeth | 9,506 | 9,506 | 1,173,746 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 786 | nemeth22 | Nemeth | 9,506 | 9,506 | 1,358,832 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 787 | nemeth23 | Nemeth | 9,506 | 9,506 | 1,506,810 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 788 | nemeth24 | Nemeth | 9,506 | 9,506 | 1,506,550 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 789 | nemeth25 | Nemeth | 9,506 | 9,506 | 1,511,758 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 790 | nemeth26 | Nemeth | 9,506 | 9,506 | 1,511,760 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 778 | nemeth14 | Nemeth | 9,506 | 9,506 | 496,144 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 779 | nemeth15 | Nemeth | 9,506 | 9,506 | 539,802 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 765 | nemeth01 | Nemeth | 9,506 | 9,506 | 725,054 | Theoretical/Quantum Chemistry Problem Sequence | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 766 | nemeth02 | Nemeth | 9,506 | 9,506 | 394,808 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 767 | nemeth03 | Nemeth | 9,506 | 9,506 | 394,808 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 768 | nemeth04 | Nemeth | 9,506 | 9,506 | 394,808 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 769 | nemeth05 | Nemeth | 9,506 | 9,506 | 394,808 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 770 | nemeth06 | Nemeth | 9,506 | 9,506 | 394,808 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |
| 771 | nemeth07 | Nemeth | 9,506 | 9,506 | 394,812 | Subsequent Theoretical/Quantum Chemistry Problem | 1999 | MATLAB Rutherford Boeing Matrix Market |