What is the next step after LU decomposition for solving Ax=b?

  • Context: Graduate 
  • Thread starter Thread starter Milentije
  • Start date Start date
  • Tags Tags
    Elimination Gauss
Click For Summary

Discussion Overview

The discussion revolves around the next steps to take after performing LU decomposition on a sparse matrix in the context of solving the linear equation Ax=b, where A is a large 1484x1484 matrix with a significant number of zero entries. Participants explore various methods and considerations for efficiently solving this system.

Discussion Character

  • Technical explanation
  • Debate/contested
  • Mathematical reasoning

Main Points Raised

  • One participant inquires about the next steps after LU decomposition for a sparse matrix, expressing uncertainty about algebraic methods.
  • Another participant suggests that simple Gauss elimination is not effective for sparse systems and recommends looking for open-source libraries for solving linear systems.
  • A participant questions the purpose of the matrix problem, indicating curiosity about its application.
  • It is proposed that the best method for solving the system may depend on the characteristics of matrix A, such as whether it is banded or symmetric, and that iterative methods might be more suitable for large, randomly distributed sparse matrices.
  • The original poster mentions the context of calculating velocities for nodes in a model, highlighting the size of the problem.
  • Another participant recommends visiting netlib for resources and suggests that various codes can be found online depending on the chosen method.

Areas of Agreement / Disagreement

Participants express differing views on the best methods to use after LU decomposition, with no consensus reached on a single approach. The discussion remains unresolved regarding the most effective strategy for the specific characteristics of the matrix.

Contextual Notes

Participants note the importance of the matrix's properties, such as symmetry and banding, which may influence the choice of method. There are also mentions of potential difficulties in installing specific libraries, which could affect the implementation of suggested methods.

Milentije
Messages
47
Reaction score
0
I have
Ax=b problem
where A 1484x1484 matrix,b 1484x1.
A is sparse(95% zeros) but if I go for LU decomposition what should be the next step?
Or is there any other method,I forgot algebra,learned it when I was undergrad long time ago.
 
Physics news on Phys.org
Simple Gauss elimination is mainly a pedagogical tool used in teaching linear systems, and it's far from being the most effective method (especially when talking about specialized problems like sparse systems). If you're writing a code that solves your linear system, you can find free open source libraries/software packages for that. Just Google "sparse linear system open source".
 
Milentije said:
I have
Ax=b problem
where A 1484x1484 matrix,b 1484x1.
A is sparse(95% zeros) but if I go for LU decomposition what should be the next step?
Or is there any other method,I forgot algebra,learned it when I was undergrad long time ago.
1484? :bugeye:

May I ask what this is for, out of curiosity?
 
The best method will depend on several factors. Is A banded and symmetric? If is narrow banded and symmetric, Gauss, Choleski decomp, or other methods may be used. If A is large, sparse, and the zero entries are somewhat randomly distributed, with no symmetry or banded layout of the non-zero terms, then an iterative method might be more suitable.
 
Yes,I am creating input file where velocities need to be calculated for every node.Total number of velocities in model is 1484,quite big.
Regarding software,I have problem to install SUPER Lu from LBNL,are there any links for simple code that to not require libraries(like BLAS( IN THIS CASE?
 
I would make my first stop at the netlib: http://www.netlib.org/

Depending on what method you use, you can always google (or dogpile) and find scads of code.
 
  • Like
Likes   Reactions: 1 person

Similar threads

  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 8 ·
Replies
8
Views
12K
  • · Replies 3 ·
Replies
3
Views
3K
Replies
1
Views
4K
  • · Replies 2 ·
Replies
2
Views
4K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 2 ·
Replies
2
Views
2K
Replies
6
Views
3K
  • · Replies 8 ·
Replies
8
Views
2K