有向图上的分布式次梯度投影算法：替代证明
最优化与控制
2017-06-26 v1
摘要
我们提出了有向-分布式投影次梯度（D-DPS）算法，以求解多智能体网络上的约束优化问题，其中智能体的目标是共同最小化局部已知凸函数之和。网络中的每个智能体仅拥有其局部目标函数，并受限于一个公共已知的凸集。我们关注智能体之间的通信由有向网络描述的情况。D-DPS 结合了盈余共识，以克服由有向通信网络引起的不对称性。分析表明收敛速率为 。
引用
@article{arxiv.1706.07707,
title = {Distributed Subgradient Projection Algorithm over Directed Graphs: Alternate Proof},
author = {Ran Xin and Chenguang Xi and Usman A. Khan},
journal= {arXiv preprint arXiv:1706.07707},
year = {2017}
}
备注
Disclaimer: This manuscript provides an alternate approach to prove the results in \textit{C. Xi and U. A. Khan, Distributed Subgradient Projection Algorithm over Directed Graphs, in IEEE Transactions on Automatic Control}. The changes, colored in blue, result into a tighter result in Theorem~1". arXiv admin note: text overlap with arXiv:1602.00653