Online Time-Varying Stochastic Quasar-Convex Optimization

This seminar from GERAD Research Center features Yuen-Man Pun from CIICADA Lab at The Australian National University discussing online dynamic stochastic optimization problems with quasar-convex loss functions. Explore how these mathematical concepts apply to real-world applications including identification of linear dynamical systems and generalized linear models. Learn about the utilization of online gradient descent and the derivation of regret bounds based on cumulative path variation and cumulative gradient variance. Discover practical applications of these theoretical concepts in generalized linear models, phase retrieval, and tomographic reconstruction. The presentation includes numerical experiments that validate the theoretical findings presented throughout the 32-minute talk scheduled for March 12, 2025.