Join Dr. Qi Wang, postdoctoral researcher in Industrial and Operations Engineering at the University of Michigan, for a seminar on advancing algorithmic methods for large-scale continuous nonlinear optimization. Drawing on her work in high-dimensional and data-intensive settings—common in modern machine learning—Dr. Wang will discuss new approaches that improve scalability, efficiency, and convergence. She will introduce an inexact trust-region method that accelerates subproblem computation in high dimensions, followed by two stochastic gradient–based algorithms for expected-value objectives with noisy gradients. These include an Adam-style method capable of handling nonconvex constraints and a line-search–based method that adaptively selects step sizes to ensure robust convergence. Together, these techniques offer principled strategies for tackling today’s most challenging large-scale optimization problems.