一个基于锥模型的自适应信赖域算法

An Adaptive Trust-region Algorithm for Unconstrained Optimization Based on the Conic Model

对无约束优化问题提出了基于锥模型的自适应信赖域算法.把锥模型子问题变成二次模型的子问题进行求解,从而减少信赖域子问题的求解.二次模型的信赖域算法是新算法的特例.在适当的条件下,证明了算法的全局收敛性及超线性收敛性.数值试验表明新算法是有效的.

This paper puts forward the adaptive trust-region algorithm based on the conic model for unconstrained optimization. The algorithm simplifies the solution of the trust-region subproblem by changing the conic-model-based subproblem to a quadrate-model subproblem. The quadratic-model algorithm is an exception to new algorithms, and under convergence and superlinear convergence. Numerical tests proper conditions, it can prove the global show the new algorithm is effective.