Abstract: In this talk, I will present some recent works on the design, analysis, and implementation of practical algorithms for solving stochastic optimization problems with constraints.
The first part of this talk focuses on the complexity theory of an inexact regularized L-shaped algorithm for two-stage stochastic programming problems. Under common assumptions including fixed recourse and bounded (sub)gradients, we provide the number of iterations, operations, and samples that the algorithm needs to find a near-optimal solution, where the radius of the convergence neighborhood depends on the level of the inexactness of objective function estimates.
In the rest of this talk, if time allows, I will introduce a step-search sequential quadratic programming method for minimizing a noisy function subject to deterministic nonlinear equality constraints. In addition to presenting the theoretical convergence behavior of such an algorithm, we compare the empirical performance of our proposed method with an objective-function-free variant that illustrates the advantages and limitations of our algorithm.
Bio: Baoyu Zhou is a postdoctoral researcher at the University of Michigan (Department of Industrial and Operations Engineering) and the University of Chicago (Booth School of Business). He received his doctoral and master’s degree in Industrial and Systems Engineering (ISE) from Lehigh University. He received his bachelor’s degree in Mechanical Engineering from Shanghai Jiao Tong University. He was a Givens Associate in the Mathematics and Computer Science Division at Argonne National Laboratory and a Research Intern at Facebook AI Research. His research focuses on developing, analyzing, and implementing practical algorithms for solving large-scale continuous optimization problems.