摘要
提出一类新的曲线搜索下的多步下降算法,在较弱条件下证明了算法具有全局收敛性和线性收敛速率.算法利用前面多步迭代点的信息和曲线搜索技巧产生新的迭代点,收敛稳定,不用计算和存储矩阵,适于求解大规模优化问题.数值试验表明算法是有效的.
This paper presents a new class of multi-step descent methods with curve search rule. We prove its global convergence and linear convergence rate under some mild conditions. The methods use to generate new iterative poin and be more suitable to solve and storage of some matrices. pr ts evious multi-step iterative information and curve search rule at each iteration. This makes the new method converge stably large scale optimization problems by avoiding the computation Numerical experiments show that the new method is available and efficient in practical computation.
出处
《应用数学》
CSCD
北大核心
2009年第4期815-820,共6页
Mathematica Applicata
基金
国家自然科学基金项目(10671166)
关键词
无约束优化
曲线搜索
全局收敛性
线性收敛速率
Unconstrained optimization
Curve search rule
Global convergence
Linear convergence rate
