PROJECT TITLE :

Playing with Duality: An overview of recent primal?dual approaches for solving large-scale optimization problems

ABSTRACT:

Optimization strategies are at the core of the many problems in signal/image processing, laptop vision, and machine learning. For a long time, it's been recognized that wanting at the dual of an optimization downside may drastically simplify its answer. But, deriving economical ways that jointly bring into play the primal and dual issues may be a a lot of recent plan that has generated several vital new contributions lately. These novel developments are grounded within the recent advances in convex analysis, discrete optimization, parallel processing, and nonsmooth optimization with an stress on sparsity issues. In this text, we aim to gift the principles of primal-twin approaches whereas providing an outline of the numerical methods that are proposed in numerous contexts. Last but not least, primal-twin strategies lead to algorithms that are simply parallelizable. These days, such parallel algorithms are changing into increasingly important for efficiently handling high-dimensional problems.