Gradient Conjugué
La méthode du gradient conjugué (CG) est une algorithme itératif used for solving systems of linear equations, specifically those that are large and sparse, which are often encountered in various scientific and engineering applications. It is particularly effective for symmetric and positive-definite matrices.
Unlike direct methods that attempt to find the exact solution in a finite number of steps, the Méthode du gradient conjugué works by generating a sequence of approximations that converge toward the solution. The key idea is to minimize the quadratic function associated with the système linéaire, which can be visualized as finding the lowest point on a surface defined by that function.
Voici comment cela fonctionne généralement :
- Initialisation : Commencez par une supposition initiale pour la solution et calculez le résidu initial.
- Processus itératif: In each iteration, update the solution by moving in a direction that is conjugate to the previous directions, which helps ensure that the search does not revisit previous paths.
- Terminaison : The process continues until the residual is sufficiently small, indicating that the solution is close enough to the true answer.
One of the main advantages of the Conjugate Gradient method is its efficiency in terms of memory and computation, especially for large systems where direct methods would be infeasible. It typically requires far fewer iterations than the number of unknowns, making it a preferred choice in various applications, including numerical simulations, optimization problems, and traitement d'image.