looking for fastest algorithm for solving banded systems of equations
From: Michael Weitzel (weitzel_at_simtec.mb.uni-siegen.de)
Date: 03/21/05
- Next message: Michael Weitzel: "Re: looking for fastest algorithm for solving banded systems of equations"
- Previous message: gcrhoads_at_yahoo.com: "Re: Where to find DFA decomposition algorithm?"
- Next in thread: Michael Weitzel: "Re: looking for fastest algorithm for solving banded systems of equations"
- Reply: Michael Weitzel: "Re: looking for fastest algorithm for solving banded systems of equations"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: 21 Mar 2005 14:43:57 +0200
Hi,
I am working with banded systems of linear equations. These systems are
characterized by matrix dimension N, a lower bandwith[1] lb and an upper
bandwidth ub (overall bandwidth is lb+1+ub).
Apart from the bandwidth and the fact that the system has full rank, there
are no other special properties (such as positive definiteness or symmetry).
* What is the running time (in terms of \Theta, asymtotic tight bound) of
the fastest algorithm from the Gauss/LU(P) algorithm-family?
* Is there a known algorithm that is faster than O(N^2) and only quadratic
in bandwidth but not in matrix dimension? ( i.e. O(N*(lb+1+ub)^2 )
Thanks for your help!
[1] The term bandwidth means, given a matrix A with Elements A(i,j),
A(i,j) = 0 where i>j+lb or j>i+ub for 1<=i,j<=N, i.e. the only
non-zero elements are in a band of width lb+1+ub around the diagonal
- Next message: Michael Weitzel: "Re: looking for fastest algorithm for solving banded systems of equations"
- Previous message: gcrhoads_at_yahoo.com: "Re: Where to find DFA decomposition algorithm?"
- Next in thread: Michael Weitzel: "Re: looking for fastest algorithm for solving banded systems of equations"
- Reply: Michael Weitzel: "Re: looking for fastest algorithm for solving banded systems of equations"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|