Abstract
In this paper, an improved complexity analysis of full Nesterov–Todd step feasible interior-point method for symmetric optimization is considered. Specifically, we establish a sharper quadratic convergence result using several new results from Euclidean Jordan algebras, which leads to a wider quadratic convergence neighbourhood of the central path for the iterates in the algorithm. Furthermore, we derive the currently best known iteration bound for full Nesterov–Todd step feasible interior-point method.
| Original language | American English |
|---|---|
| Journal | Journal of Optimization Theory and Applications |
| Volume | 166 |
| DOIs | |
| State | Published - Aug 1 2015 |
Disciplines
- Education
- Mathematics
Keywords
- Euclidean Jordan algebras
- Full Nesterov–Todd step
- Interior-point methods
- Linear optimization over symmetric cones
- Polynomial complexity