scientific article; zbMATH DE number 4028249
[For the entire collection see Zbl 0626.00020.] We introduce two metrics between closed convex-concave functions. The first is defined in terms of inf-sup convolutive approximations and the second relies on the associated resolvent operators. The partial Legendre-Fenchel transform is shown to be an isometry relative to the first distance. From the comparison between these two metrics, it appears that the saddle-points of the coercified Lagrangians converge at a rate which is equal to the square root of the distance between the convex- concave functions. An application is given to convex programming.
- scientific article; zbMATH DE number 25334
- scientific article; zbMATH DE number 25342
- Convergence of convex-concave saddle functions: Applications to convex programming and mechanics
- scientific article; zbMATH DE number 4119062
- A two-level subgradient method for finding saddle points of convex- concave functions
- Convergence of convex-concave saddle functions: Applications to convex programming and mechanics
- Topologies and convergences on the space of convex functions
- scientific article; zbMATH DE number 4119062 (Why is no real title available?)
- About saddle values of a class of convex-concave functions
- The landscape of the proximal point method for nonconvex-nonconcave minimax optimization
- Approximations of Rockafellians, Lagrangians, and dual functions
- The rate of convergence of Sinkhorn balancing
This page was built for publication:
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3769494)