Associate Professor Xiao Wang, Pengcheng Laboratory

Nonconvex constrained optimization (NCO) has been one of the important research fields in optimization community. It has widely appeared in many application fields. However, challenges for solving NCO often arise due to larger scale of data and uncertainty involved in optimization models. In this talk, I will briefly introduce our recent progress on stochastic approximation (SA) methods for NCO, including stochastic primal-dual methods for problems with a large number of constraints and a nest structure, respectively, a zeroth-order SA method for NCO to explore the complexity dependence on dimensionarity and a SA method for nonconvex equality constrained optimization to pursue second-order stationarity.