Presentation is loading. Please wait.

Presentation is loading. Please wait.

Conjugate gradient iteration One matrix-vector multiplication per iteration Two vector dot products per iteration Four n-vectors of working storage x 0.

Similar presentations


Presentation on theme: "Conjugate gradient iteration One matrix-vector multiplication per iteration Two vector dot products per iteration Four n-vectors of working storage x 0."— Presentation transcript:

1 Conjugate gradient iteration One matrix-vector multiplication per iteration Two vector dot products per iteration Four n-vectors of working storage x 0 = 0, r 0 = b, d 0 = r 0 for k = 1, 2, 3,... α k = (r T k-1 r k-1 ) / (d T k-1 Ad k-1 ) step length x k = x k-1 + α k d k-1 approx solution r k = r k-1 – α k Ad k-1 residual β k = (r T k r k ) / (r T k-1 r k-1 ) improvement d k = r k + β k d k-1 search direction

2 Conjugate gradient: Convergence In exact arithmetic, CG converges in n steps (completely unrealistic!!) Accuracy after k steps of CG is related to: consider polynomials of degree k that are equal to 1 at 0. how small can such a polynomial be at all the eigenvalues of A? Thus, eigenvalues close together are good. Condition number: κ (A) = ||A|| 2 ||A -1 || 2 = λ max (A) / λ min (A) Residual is reduced by a constant factor by O(κ 1/2 (A)) iterations of CG.

3 The Landscape of Sparse Ax=b Solvers Direct A = LU Iterative y’ = Ay Non- symmetric Symmetric positive definite More RobustLess Storage More Robust More General D

4 Other Krylov subspace methods Nonsymmetric linear systems: GMRES: for i = 1, 2, 3,... find x i  K i (A, b) that minimizes ||b – Ax i || 2 But, no short recurrence => save old vectors => lots more space (Usually “restarted” every k iterations to use less space.) BiCGStab, QMR, etc.: Two spaces K i (A, b) and K i (A T, b) w/ mutually orthogonal bases Short recurrences => O(n) space, but less robust Convergence and preconditioning more delicate than CG Active area of current research Eigenvalues: Lanczos (symmetric), Arnoldi (nonsymmetric)

5 Preconditioners Suppose you had a matrix B such that: 1.condition number κ (B -1 A) is small 2.By = z is easy to solve Then you could solve (B -1 A)x = B -1 b instead of Ax = b Actually (B -1/2 AB -1/2 ) (B 1/2 x) = B -1/2 b, but never mind… B = A is great for (1), not for (2) B = I is great for (2), not for (1) Domain-specific approximations sometimes work B = diagonal of A sometimes works Better: blend in some direct-methods ideas...

6 Preconditioned conjugate gradient iteration x 0 = 0, r 0 = b, d 0 = B -1 r 0, y 0 = B -1 r 0 for k = 1, 2, 3,... α k = (y T k-1 r k-1 ) / (d T k-1 Ad k-1 ) step length x k = x k-1 + α k d k-1 approx solution r k = r k-1 – α k Ad k-1 residual y k = B -1 r k preconditioning solve β k = (y T k r k ) / (y T k-1 r k-1 ) improvement d k = y k + β k d k-1 search direction One matrix-vector multiplication per iteration One solve with preconditioner per iteration

7 Incomplete Cholesky factorization (IC, ILU) Compute factors of A by Gaussian elimination, but ignore fill Preconditioner B = R T R  A, not formed explicitly Compute B -1 z by triangular solves (in time nnz(A)) Total storage is O(nnz(A)), static data structure Either symmetric (IC) or nonsymmetric (ILU) x ARTRT R

8 Incomplete Cholesky and ILU: Variants Allow one or more “levels of fill” unpredictable storage requirements Allow fill whose magnitude exceeds a “drop tolerance” may get better approximate factors than levels of fill unpredictable storage requirements choice of tolerance is ad hoc Partial pivoting (for nonsymmetric A) “Modified ILU” (MIC): Add dropped fill to diagonal of U or R A and R T R have same row sums good in some PDE contexts 2 3 1 4 2 3 1 4

9 Incomplete Cholesky and ILU: Issues Choice of parameters good: smooth transition from iterative to direct methods bad: very ad hoc, problem-dependent tradeoff: time per iteration (more fill => more time) vs # of iterations (more fill => fewer iters) Effectiveness condition number usually improves (only) by constant factor (except MIC for some problems from PDEs) still, often good when tuned for a particular class of problems Parallelism Triangular solves are not very parallel Reordering for parallel triangular solve by graph coloring

10 Matrix permutations for iterative methods Symmetric matrix permutations don’t change the convergence of unpreconditioned CG Symmetric matrix permutations affect the quality of an incomplete factorization – poorly understood, controversial Often banded (RCM) is best for IC(0) / ILU(0) Often min degree & nested dissection are bad for no-fill incomplete factorization but good if some fill is allowed Some experiments with orderings that use matrix values e.g. “minimum discarded fill” order sometimes effective but expensive to compute

11 Sparse approximate inverses Compute B -1  A explicitly Minimize || A B -1 – I || F (in parallel, by columns) Variants: factored form of B -1, more fill,.. Good: very parallel, seldom breaks down Bad: effectiveness varies widely AB -1

12 Nonsymmetric matrix permutations for iterative methods Dynamic permutation: ILU with row or column pivoting E.g. ILUTP (Saad), Matlab “luinc” More robust but more expensive than ILUT Static permutation: Try to increase diagonal dominance E.g. bipartite weighted matching (Duff & Koster) Often effective; no theory for quality of preconditioner Field is not well understood, to say the least

13 Row permutation for heavy diagonal Row permutation for heavy diagonal [Duff, Koster] Represent A as a weighted, undirected bipartite graph (one node for each row and one node for each column) Find matching (set of independent edges) with maximum product of weights Permute rows to place matching on diagonal Matching algorithm also gives a row and column scaling to make all diag elts =1 and all off-diag elts <=1 15234 1 5 2 3 4 A 1 5 2 3 4 1 5 2 3 4 15234 4 2 5 3 1 PA

14 Complexity of direct methods n 1/2 n 1/3 2D3D Space (fill): O(n log n)O(n 4/3 ) Time (flops): O(n 3/2 )O(n 2 ) Time and space to solve any problem on any well- shaped finite element mesh

15 Complexity of linear solvers 2D3D Dense Cholesky:O(n 3 ) Sparse Cholesky:O(n 1.5 )O(n 2 ) CG, exact arithmetic: O(n 2 ) CG, no precond: O(n 1.5 )O(n 1.33 ) CG, modified IC0: O(n 1.25 )O(n 1.17 ) CG, support trees: O(n 1.20 ) -> O(n 1+ )O(n 1.75 ) -> O(n 1+ ) Multigrid:O(n) n 1/2 n 1/3 Time to solve model problem (Poisson’s equation) on regular mesh


Download ppt "Conjugate gradient iteration One matrix-vector multiplication per iteration Two vector dot products per iteration Four n-vectors of working storage x 0."

Similar presentations


Ads by Google