Speaker：Chunming Tang（Guangxi University）

Time：2022-12-23, 16:00

Location：Tencent Meeting ID：807-394-121（No Password）

Abstract:

In this talk, a restricted memory quasi-Newton bundle method for minimizing a locally Lipschitz function over a Riemannian manifold is proposed, which extends the classical ones in Euclidean space to the manifold setting. The potential second order information of the objective function is approximated by applying the Riemannian versions of the quasi-Newton updating formulas. The subgradient aggregation technique is used to avoid solving the time-consuming quadratic programming subproblem when calculating the candidate descent direction. Moreover, a new Riemannian line search procedure is proposed to generate the stepsizes. Global convergence of the proposed method is established: if the serious iteration steps are finite, then the last serious iteration is stationary; otherwise every accumulation point of the serious iteration sequence is stationary. Finally, some preliminary numerical results show that the proposed method is promising.