报告主题：Parallelizable second-order approach for optimization problems with orthogonality constraints
报告人：刘歆 副研究员 （中国科学院数学与系统科学研究院）
报告时间：2019年6月14日（周五）9：00
报告地点：校本部G507
邀请人：白延琴教授
报告摘要：Updating the augmented Lagrangian multiplier by closed-form expression yields efficient infeasible approach for optimization problems with orthogonality constraints. Hence, parallelization becomes tractable in solving this type of problems. To accelerate the local convergence, we consider second-order approach under this framework. To avoid expensive calculation or solving a hard subproblem in computing the Newton step, we propose a new strategy to do it approximately which leads to superlinear convergence theoretically. In practice, the new second-order approach outperforms the existent algorithms. Last but not least, this new approach is completely orthonormalization-free and hence can be parallelized directly.
欢迎教师、学生参加！