A RESIDUAL-NORM BASED ALGORITHM FOR SOLVING DETERMINATE/INDETERMINATE SYSTEMS OF NON-LINEAR ALGEBRAIC EQUATIONS

摘要

In this paper, a residual-norm based algorithm (RNBA) is applied to solve determinate/indeterminate systems of nonlinear algebraic equations. The RNBA is derived from a manifold and defined in terms of a squared residual norm and a fictitious time variable, from which a robust iterative algorithm with either fixing or adjusting parameters can be obtained. Besides, some convergent indexes, such as manifold factor AO and ratio of residual errors S, are defined to indicate the manifold attracting effect. Through the convergent indexes, the convergent mechanism of the RNBA displays a Hopf bifurcation when approximating the true solutions. Several numerical examples, including the root finding in two-, three-variable and in elliptic-type partial differential equations (PDEs), are examined. Comparisons of numerical results and exact solutions show that the proposed algorithm has good computational efficiency and accuracy simultaneously.