数学与系统科学研究院

计算数学所学术报告

报告人：        Wang Zhouhong

Institute of Computational Mathematics and Scientific/Engineering Computing , CAS

报告题目 Subspace Properties of Trust Region Method for Unconstrained Optimization

Abstract: In this talk we will talk about the subspace properties of trust region methods for unconstrained optimization, assuming the approximate Hessian is updated by quasi-Newton formulae. It is shown that the trial step obtained by solving the trust region subproblem is in the subspace spanned by all the gradient vectors computed. Thus, the trial step can also be defined by minimizing the quasi-Newton quadratic model in the subspace. Based on this observation, a subspace trust region algorithm is proposed, and we can solve a smaller trust region subproblem in the algorithm, which would reduce the computations, particularly for early iterations. Then a subspace implementation method for limited memory trust region methods is discussed. Numerical results are also reported.

报告时间：2004年9月22日  下午4：00-5：00

报告地点：科技综合楼三层报告厅