无约束优化问题的循环梯度法

2022.12.13

投稿:龚惠英部门:浏览次数:

活动信息

报告题目 (Title):Cyclic gradient methods for unconstrained optimization(无约束优化问题的循环梯度法)

报告人 (Speaker): 孙聪 副教授(北京邮电大学)

报告时间 (Time):2022年12月13日(周二) 10:00-11:00

报告地点 (Place):腾讯会议 (会议号:225-921-874)

邀请人(Inviter):徐姿

主办部门:理学院数学系

报告摘要:In this work, the cyclic gradient method with Cauchy steps and constructed constant steps for convex quadratic function minimization problems are extended to general smooth unconstrained optimization problems. The Cauchy steps are approximated first by the step-ahead BB steps, and then by the interpolation. The convergence properties of the proposed methods are analyzed. Numerical results show the good performances of the proposed gradient methods compared to the benchmarks.