摘要
以牛顿法为基础 ,通过符号运算求导和信赖域方法 ,解决其Hesse矩阵计算工作量大和局部收敛性的问题 ,设计和实现了非线性最小二乘的通用算法。该方法计算速度快 ,计算精度高 ,对初始值的选择不敏感 ,不仅可以直接用于线性最小二乘 ,而且可以适用于大数据量的非线性最小二乘。
Based on Newton algorithm, using symbol calculation and trust region method for solving huge workload of Hesse matrix and local convergence of Newton algorithm, a general algorithm has been designed and implemented. Because of its high computation speed, high precision and non sensitive to initial value, it is not only used directly for linear least square, but it benefits nonlinear least square of huge data. The numeric test shows the feasibility of the algorithm.
出处
《计算机应用》
CSCD
北大核心
2004年第7期22-24,共3页
journal of Computer Applications
基金
国家自然科学基金资助项目 (30 2 71 0 79)
国家林业局 948项目 (2 0 0 1 - 1 3)
关键词
符号运算
信赖域方法
非线性最小二乘
symbol calculation
trust region method
nonlinear least square
