数学与系统科学研究院

计算数学所学术报告

报告人：        袁亚湘  研究员

                              中国科学院计算数学所

                                yyx@lsec.cc.ac.cn

报告题目    Subspace Techniques in Trust Region Algorithms for Nonlinear Optimization

Abstract: Trust region algorithms are a class of numerical algorithms for nonlinear optimization. In this talk we will review the recent advances in trust region algorithms, partically the subspace techniques. The use of subspaces can enable us to construct small size subproblems even thought the original problem is very large. The subspace approach can also be viewed as an improvement and generalization of the limited memory quasi-Newton technique. For constrained optimization problems, trial steps of trust region algorithms can be decomposited into null space steps and range space steps. The theoretical properties of these steps are also discussed using the subspace techniques.

报告时间：2003年9月4日  下午4：00-5：00

报告地点：科技综合楼三层报告厅