发表机构
Purdue University; Mila - Quebec AI Institute; McGill University(普渡大学; 米拉-魁北克人工智能研究所; 麦吉尔大学)
机构由 AI 辅助整理,请以论文原文为准。AI 中文总结
本文提出Dec-BFTRL算法,用于分离访问下分散式在线上线性化优化,实现平方根遗憾,并应用于连续子模最大化。
AI 中文摘要
我们研究了在高效分离访问下,动作集上上线性化收益的分散式在线优化问题,并将其应用于在线连续递减收益(DR)子模最大化。我们提出了分散式障碍跟随正则化领导者(Dec-BFTRL)算法,并将每个智能体采取的动作与所有局部目标的平均值进行比较。每个智能体通过近似规范投影将内部迭代映射到可行动作,仅通信累积的代理梯度对偶状态,并调用局部混合牛顿过程来近似最小化其通信后的BFTRL势函数。对于每个智能体,我们实现了期望的网络聚合遗憾为$\widetilde O(\sqrt{T})$。在$T$轮中,每个智能体使用$T$次邻居混合步骤和$\widetilde O(T)$次分离预言机调用。我们给出了四个包装实例,涵盖三个DR子模最大化问题。
英文摘要
We study decentralized online optimization of upper-linearizable payoffs over an action set under efficient separation access, with applications to online continuous diminishing-return (DR) submodular maximization. We propose Decentralized Barrier Follow-the-Regularized-Leader (Dec-BFTRL), and evaluate each agent's played action against the average of all local objectives. Each agent maps an internal iterate to a feasible action through an approximate gauge projection, communicates only a cumulative surrogate-gradient dual state, and invokes the local HybridNewton procedure to approximately minimize its post-communication BFTRL potential. For every agent, we achieve expected network-aggregate regret of $\widetilde O(\sqrt{T})$. Over $T$ rounds, each agent uses $T$ neighbor-mixing steps and $\widetilde O(T)$ separation-oracle calls. We give wrapper instantiations covering four up-concave or DR-submodular maximization problems.