On Glowinski’s Open Question of Alternating Direction Method of Multipliers
نویسندگان
چکیده
The alternating direction method of multipliers (ADMM) was proposed by Glowinski and Marrocco in 1975; and it has been widely used in a broad spectrum of areas, especially in some sparsitydriven application domains. In 1982, Fortin and Glowinski suggested to enlarge the range of the step size for updating the dual variable in ADMM from 1 to (0, 1+ √ 5 2 ); and this strategy immediately accelerates the convergence of ADMM for most of its applications. Meanwhile, Glowinski raised the question of whether or not the range can be further enlarged to (0, 2); this question remains open with nearly no progress in the past decades. In this paper, we answer this question affirmatively for the case where both the functions in the objective are quadratic. Glowinski’s open question is thus partially answered. We further establish the global linear convergence of the ADMM with the step size range (0, 2) for the quadratic programming case under a condition that turns out to be tight.
منابع مشابه
Modified Convex Data Clustering Algorithm Based on Alternating Direction Method of Multipliers
Knowing the fact that the main weakness of the most standard methods including k-means and hierarchical data clustering is their sensitivity to initialization and trapping to local minima, this paper proposes a modification of convex data clustering in which there is no need to be peculiar about how to select initial values. Due to properly converting the task of optimization to an equivalent...
متن کاملOn the Step Size of Symmetric Alternating Directions Method of Multipliers
The alternating direction method of multipliers (ADMM) is an application of the Douglas-Rachford splitting method; and the symmetric version of ADMM which updates the Lagrange multiplier twice at each iteration is an application of the Peaceman-Rachford splitting method. Sometimes the symmetric ADMM works empirically; but theoretically its convergence is not guaranteed. It was recently found th...
متن کاملOn the O(1/t) Convergence Rate of Alternating Direction Method with Logarithmic-Quadratic Proximal Regularization
It was recently shown that the alternating direction method with logarithmicquadratic proximal regularization can yield an efficient algorithm for a class of variational inequalities with separable structures. This paper further shows the O(1/t) convergence rate for this kind of algorithms. Both the cases with a simple or general Glowinski’s relaxation factor are discussed.
متن کاملOn the O(1/t) convergence rate of Eckstein and Bertsekas’s generalized alternating direction method of multipliers
This note shows the O(1/t) convergence rate of Eckstein and Bertsekas’s generalized alternating direction method of multipliers in the context of convex minimization with linear constraints.
متن کاملDistributed Learning-to-Rank on Streaming Data using Alternating Direction Method of Multipliers
We show that Alternating Direction Method of Multipliers is an effective method for largescale learning-to-rank on multi-cores and clusters, especially in scenarios requiring joint distributed and streaming architectures.
متن کامل