For unconstrained optimization problems,quasi-Newton methods are a class of utilizing,first derivative,the most effective approach.How to improve the computing efficiency of the actual calculation,and how to obtain global convergence and local super linear convergence for non-convex objective function are contents of quasi-Newton research.This paper compares some modified quasi-Newton methods in the recent literature,proposes a modified BFGS quasi-Newton method and analyzes its global convergence.%拟牛顿法是所有利用一阶导数求解无约束优化问题的方法中最有效的一类计算方法,如何提高实际计算中的运算效率,如何使得对非凸目标函数保持局部超线性收敛的同时具有全局收敛性,是对拟牛顿法进行研究的两个方向.对近年来相关文献的几种修正拟牛顿法进行分析比较,并提出和分析了一个修正BFGS拟牛顿法的收敛性.
展开▼