A New Nonmonotone Memory Gradient Method for Unconstrained Optimization
Yuzhu Li, Qinghua Zhou*
College of Mathematics and Information Science, Hebei University, Baoding, 071002, China
Abstract: In this paper, we propose and analyze a new non-monotone memory gradient method for unconstrained optimization. Actually, we combine a non-monotone strategy into a modified Armijo rule and design a new non-monotone memory gradient algorithm that possibly chooses a larger step length. Generally, the global convergence is analyzed under some suitable conditions.
Keywords: Unconstrained optimization; Memory gradient methods; Non-monotone strategy; Global convergence.