Adaptive Consensus ADMM for Distributed Optimization
About
The alternating direction method of multipliers (ADMM) is commonly used for distributed model fitting problems, but its performance and reliability depend strongly on user-defined penalty parameters. We study distributed ADMM methods that boost performance by using different fine-tuned algorithm parameters on each worker node. We present a O(1/k) convergence rate for adaptive ADMM methods with node-specific parameters, and propose adaptive consensus ADMM (ACADMM), which automatically tunes parameters without user oversight.
Zheng Xu, Gavin Taylor, Hao Li, Mario Figueiredo, Xiaoming Yuan, Tom Goldstein• 2017
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Elastic net regression | Synthetic Elastic net regression | Runtime (s)623 | 69 | |
| Elastic net regression | synthetic 2 | Iterations57 | 5 | |
| Elastic net regression | MNIST | Iterations14 | 5 | |
| Elastic net regression | news20 | Iterations78 | 5 | |
| Elastic net regression | RCV1 | Iterations8 | 5 | |
| Elastic net regression | Realsim | Iterations9 | 5 | |
| Semidefinite programming | Ham-9-5-6 | Iterations30 | 5 | |
| Sparse Logistic Regression | synthetic 1 | Iterations24 | 5 | |
| Sparse Logistic Regression | synthetic 2 | Iterations114 | 5 | |
| Sparse Logistic Regression | MNIST | Iterations149 | 5 |
Showing 10 of 22 rows