科研之友
微信
新浪微博
Facebook
分享链接
摘要
We extend the definition of sufficient descent direction to the setting of Riemannian geometry. We propose an algorithm based on sufficient descent directions for solving the problem of minimizing quasiconvex functions over a Riemannian manifold. The choice of the stepsize is given by Armijo search with backtracking procedure. We show full convergence by considering complete finite-dimensional Riemannian manifolds with nonnegative sectional curvature. Computational experiments illustrate the behaviour of the method and its comparison with the steepest descent method is also given.
- 出版日期2012-10
