Nonmonotone Forward-Backward Splitting Method for a Class of Infinite-Dimensional Nonsmooth Nonconvex Problems

Behzad Azmi (U Konstanz)

Jun 04. 2024, 15:30 — 16:00

This talk focuses on the convergence analysis of a nonmonotone forward-backward splitting method for tackling a class of nonsmooth composite problems in Hilbert spaces. The objective function is the sum of a Fréchet differentiable (not necessarily convex) function and a lower semicontinuous convex function. These problems are commonly encountered in optimization problems involving nonlinear partial differential equations with sparsity-promoting cost functionals. We discuss the convergence and complexity of the algorithm. In particular, linear convergence will be derived under quadratic growth-type conditions. Additionally, we provide insights from numerical experiments that validate our theoretical findings.

 

Further Information
Venue:
ESI Boltzmann Lecture Hall
Recordings:
Recording
Associated Event:
One World Optimization Seminar in Vienna (Workshop)
Organizer(s):
Radu Ioan Bot (U of Vienna)
Yurii Malitskyi (U of Vienna)