TOPICS
Search

Preconditioner


A preconditioner for a linear system Ax=b is an invertible matrix M for which systems with M are substantially easier to solve than systems with A, while M^(-1)A has more favorable spectral or conditioning properties. Left preconditioning replaces the system by

 M^(-1)Ax=M^(-1)b,

and right preconditioning instead solves AM^(-1)y=b with x=M^(-1)y.

Effective preconditioning can greatly reduce the iterations required by a conjugate gradient method or another Krylov-subspace method.


See also

Conjugate Gradient Method, Condition Number, Linear System of Equations

Explore with Wolfram|Alpha

References

Saad, Y. Iterative Methods for Sparse Linear Systems, 2nd ed. Philadelphia, PA: SIAM, 2003. https://doi.org/10.1137/1.9780898718003.

Cite this as:

Weisstein, Eric W. "Preconditioner." From MathWorld--A Wolfram Resource. https://mathworld.wolfram.com/Preconditioner.html

Subject classifications