Tutte colloquium-David Torregrossa Belén
Title:Splitting algorithms for monotone inclusions with minimaldimension
Speaker: | David Torregrossa Belén |
Affiliation: | Center forMathematical Modeling, University of Chile |
Location: | MC 5501 |
Abstract: Many situations in convex optimization can be modeled as the problem offinding a zero of a monotone operator, which can be regarded as ageneralization of the gradient of a differentiable convex function. Inorder to numerically address this monotone inclusion problem, it isvital to be able to exploit the inherent structure of the operatordefining it. The algorithms in the family of the splitting methodsachieve this by iteratively solving simpler subtasks that are defined byseparately using some parts of the original problem.
In the first part of this talk, we will introduce some of the mostrelevant monotone inclusion problems and present their applications tooptimization. Subsequently, we will draw our attention to a commonanomaly that has persisted in the design of methods in this family: thedimension of the underlying space —which we denote as lifting— of thealgorithms abnormally increases as the problem size grows. This hasdirect implications on the computational performance of the method as aresult of the increase of memory requirements. In this framework, wecharacterize the minimal lifting that can be obtained by splittingalgorithms adept at solving certain general monotone inclusions.Moreover, we present splitting methods matching these lifting bounds, and thushaving minimal lifting.