The Minimizer of the Sum of Two Strongly Convex Functions
arXiv:2305.13134 · doi:10.1080/02331934.2024.2402923
Abstract
The optimization problem concerning the determination of the minimizer for the sum of convex functions holds significant importance in the realm of distributed and decentralized optimization. In scenarios where full knowledge of the functions is not available, limiting information to individual minimizers and convexity parameters -- either due to privacy concerns or the nature of solution analysis -- necessitates an exploration of the region encompassing potential minimizers based solely on these known quantities. The characterization of this region becomes notably intricate when dealing with multivariate strongly convex functions compared to the univariate case. This paper contributes outer and inner approximations for the region harboring the minimizer of the sum of two strongly convex functions, given a constraint on the norm of the gradient at the minimizer of the sum. Notably, we explicitly delineate the boundaries and interiors of both the outer and inner approximations. Intriguingly, the boundaries as well as the interiors turn out to be identical. Furthermore, we establish that the boundary of the region containing potential minimizers aligns with that of the outer and inner approximations.
arXiv comment41 pages, 5 figures
References in corpus (7)
- Guaranteed Minimum-Rank Solutions of Linear Matrix Equations via Nuclear Norm Minimization
- Dual Averaging for Distributed Optimization: Convergence Analysis and Network Scaling
- Non-convex Optimization for Machine Learning
- Distributed Recursive Least-Squares: Stability and Performance Analysis
- Augmented L1 and Nuclear-Norm Models with a Globally Linearly Convergent Algorithm
- ByRDiE: Byzantine-resilient distributed coordinate descent for decentralized learning
- Scalable Distributed Optimization of Multi-Dimensional Functions Despite Byzantine Adversaries