An Adaptive Primal-Dual Subgradient Algorithm for Online Distributed Constrained Optimization

Deming Yuan,Daniel W. C. Ho,Guo-Ping Jiang
DOI: https://doi.org/10.1109/tcyb.2017.2755720
IF: 11.8
2018-01-01
IEEE Transactions on Cybernetics
Abstract:In this paper, we consider the problem of solving distributed constrained optimization over a multiagent network that consists of multiple interacting nodes in online setting, where the objective functions of nodes are time-varying and the constraint set is characterized by an inequality. Through introducing a regularized convex-concave function, we present a consensus-based adaptive primal-dual subgradient algorithm that removes the need for knowing the total number of iterations T in advance. We show that the proposed algorithm attains an O(T1/2+c) [where c is an element of (0, 1/ 2)] regret bound and an O(T1-c/2) bound on the violation of constraints; in addition, we show an improvement to an O(T-c) regret bound when the objective functions are strongly convex. The proposed algorithm allows a novel trade-offs between the regret and the violation of constraints. Finally, a numerical example is provided to illustrate the effectiveness of the algorithm.
What problem does this paper attempt to address?