Corpus ID: 73528173

A General Convergence Result for Mirror Descent with Armijo Line Search

@article{Li2018AGC,
  title={A General Convergence Result for Mirror Descent with Armijo Line Search},
  author={Yen-Huan Li and C. Riofr'io and V. Cevher},
  journal={arXiv: Optimization and Control},
  year={2018}
}
Existing convergence guarantees for the mirror descent algorithm require the objective function to have a bounded gradient or be smooth relative to a Legendre function. The bounded gradient and relative smoothness conditions, however, may not hold in important applications, such as quantum state tomography and portfolio selection. In this paper, we propose a local version of the relative smoothness condition as a generalization of its existing global version, and prove that under this local… Expand
2 Citations

Figures from this paper

Online Positron Emission Tomography By Online Portfolio Selection
  • Yen-Huan Li
  • Computer Science
  • ICASSP 2020 - 2020 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
  • 2020

References

SHOWING 1-10 OF 40 REFERENCES
Mirror descent and nonlinear projected subgradient methods for convex optimization
Interior Gradient and Proximal Methods for Convex and Conic Optimization
Variable Metric Inexact Line-Search-Based Methods for Nonsmooth Optimization
On the Global Linear Convergence of Frank-Wolfe Optimization Variants
A Descent Lemma Beyond Lipschitz Gradient Continuity: First-Order Methods Revisited and Applications
On the Goldstein-Levitin-Polyak gradient projection method
Frank-Wolfe works for non-Lipschitz continuous gradient objectives: Scalable poisson phase retrieval
  • G. Ódor, Yen-Huan Li, +4 authors V. Cevher
  • Mathematics, Computer Science
  • 2016 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
  • 2016
Convex Optimization: Algorithms and Complexity
...
1
2
3
4
...