Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Adaptive ADMM with Spectral Penalty Parameter Selection

About

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.

Zheng Xu, Mario A. T. Figueiredo, Tom Goldstein• 2016

Related benchmarks

TaskDatasetResultRank
Elastic net regressionSynthetic Elastic net regression
Runtime (s)0.046
69
Sparse CT Image Reconstructionl1 fidelity TV reg
Relative Residual0.496
14
Constrained Quadratic OptimizationScaled Quads m = 2
Relative Residual0.409
14
Constrained Quadratic OptimizationScaled Quads m = 0
Relative Residual1.52e-5
14
Constrained Quadratic OptimizationScaled Quads m = 1
Relative Residual0.0074
7
Sum of quadratics with multiple constraints optimizationScaled Quads m = 1
Median Relative Residual0.0051
7
Constrained Quadratic OptimizationComplex Quads
Relative residual8.29e-9
7
Support Vector Machinesynthetic 1
Iterations19
5
Elastic net regressionMNIST
Iterations40
5
Elastic net regressionRCV1
Iterations20
5
Showing 10 of 51 rows

Other info

Follow for update