SIPo: From Continuous Dynamics to Optimization Algorithms

Abstract: 

Optimization algorithms are essentially discrete iterative procedures. However, many of them can be derived from continuous dynamical systems through suitable discretization schemes. In this talk, we explore this perspective in the context of primal–dual optimization methods for constrained and saddle-point problems. We introduce the idea of proximal operators, explain how different discretizations lead to different algorithms, and compare the resulting methods in numerical instances.

Speaker: Juan José Maulén (CMM, U. de Chile)