Adaptive ADMM with Spectral Penalty Parameter Selection
arXiv:1605.07246
Abstract
The alternating direction method of multipliers (ADMM) is a versatile tool for solving a wide range of constrained optimization problems, with differentiable or non-differentiable objective functions. Unfortunately, its performance is highly sensitive to a penalty parameter, which makes ADMM often unreliable and hard to automate for a non-expert user. We tackle this weakness of ADMM by proposing a method to adaptively tune the penalty parameters to achieve fast convergence. The resulting adaptive ADMM (AADMM) algorithm, inspired by the successful Barzilai-Borwein spectral method for gradient descent, yields fast convergence and relative insensitivity to the initial stepsize and problem scaling.
AISTATS 2017
References in corpus (13)
- Discussion of "Least angle regression" by Efron et al
- Rejoinder to "Least angle regression" by Efron et al
- Least Angle Regression
- Discussion of "Least angle regression" by Efron et al
- Discussion of "Least angle regression" by Efron et al
- Discussion of "Least angle regression" by Efron et al
- Discussion of "Least angle regression" by Efron et al
- Discussion of "Least angle regression" by Efron et al
- Discussion of "Least angle regression" by Efron et al
- Discussion of "Least angle regression" by Efron et al
- Linearized Alternating Direction Method with Adaptive Penalty for Low-Rank Representation
- A Field Guide to Forward-Backward Splitting with a FASTA Implementation
- A General Analysis of the Convergence of ADMM
Cited by in corpus (9)
- Improving the Privacy and Accuracy of ADMM-Based Distributed Algorithms
- An Empirical Study of ADMM for Nonconvex Problems
- How is Distributed ADMM Affected by Network Topology?
- Data Multiplexing in Radio Interferometric Calibration
- Newton-ADMM: A Distributed GPU-Accelerated Optimizer for Multiclass Classification Problems
- Flexible MPC-based Conflict Resolution Using Online Adaptive ADMM
- Recycled ADMM: Improve Privacy and Accuracy with Less Computation in Distributed Algorithms
- Non-negative Factorization of the Occurrence Tensor from Financial Contracts
- Improving Rate of Convergence via Gain Adaptation in Multi-Agent Distributed ADMM Framework