pipette
ENEnglish

Inertial parameterized Douglas-Rachford for generalized DC programming problems

Avinash Dixit, Pankaj Gautam, Sorin-Mihai Grad

PreprintAfirmaciones fuertes, leer con cuidado

En palabras de los autores

We propose two new Douglas-Rachford type algorithms for solving generalized DC programming problems. More precisely, we deal with the minimization of the sum of two possibly nonsmooth convex functions from which a Fr\'echet differentiable convex function with a Lipschitz continuous gradient is subtracted, in a Hilbertian framework. We introduce an inertial Douglas-Rachford DC algorithm and a parametrized inertial Douglas-Rachford DC algorithm for solving optimization problems of this type, and study their convergence behavior. Numerical experiments illustrate the behavior of the proposed algorithms when solving sparse recovery and robust classification problems, highlighting their superior performance in comparison with their state-of-the-art counterparts.

Resultado principalEl resumen no menciona limitaciones.

Apareció: viernes, 9 de octubre. arXiv. Preprint, todavía sin revisión por pares.