ADMM

Uses

How it works

Highlighted claims

• ADMM is used for structured convex optimization because it has inexpensive iterations and scales across several application areas. — Learning Over-Relaxation Policies for ADMM with Convergence Guarantees
• The paper studies relaxed ADMM for convex problems of the form minimizing f(x) + g(z) subject to a linear constraint. — Learning Over-Relaxation Policies for ADMM with Convergence Guarantees
• The paper replaces scalar rho and alpha with diagonal matrices that can vary by constraint and by iteration. — Learning Over-Relaxation Policies for ADMM with Convergence Guarantees
• Relaxed ADMM replaces part of the update with a combination controlled by a relaxation parameter alpha. — Learning Over-Relaxation Policies for ADMM with Convergence Guarantees