---
格式版本: 2
标题: "Dec-BFTRL: Squre-Root Regret for Decentralized Online Upper-Linearizable Optimization under Separation Access with Application to Continuous Submodular Maximization"
原文链接: "https://arxiv.org/abs/2608.30271"
发布日期: "2026-08-31"
发布时间校准状态: "found"
发布时间需复核: "否"
发布时间来源: "rule:local:strict_original_body"
发布时间证据: "**\\[v1\\]** Mon, 31 Aug 2026 05:29:54 UTC (40 KB)"
发布时间校准原因: "规则确认唯一严格发布时间，来源 local:strict_original_body"
发布时间校准置信度: "high"
发布时间候选数量: 18
发布时间严格候选数量: 6
发布时间原页读取状态: "source template page reused from URL open"
发布时间未找到原因: ""
发布时间校准时间: "2026-09-02T03:56:52+08:00"
发布时间仲裁状态: "skipped"
发布时间仲裁尝试次数: 0
发布时间仲裁耗时毫秒: 0
发现时间: "2026-09-02T03:54:32+08:00"
入库时间: "2026-09-01T19:56:52.283Z"
来源平台: "arXiv 学术论文搜索"
搜索渠道: "source_template"
搜索词: "https://arxiv.org/search/?query=Oracle&searchtype=all"
匹配关键词:
  - "AI"
相关厂家:
  - "Oracle"
相关专家:
  []
内容类型: "网页"
抓取工具: "Free Fetch + Defuddle"
清洗工具: "Defuddle Markdown + Defuddle/Readability 正文提取"
原始附件:
  []
AI优质: "否"
AI打分: 8
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "该页面为arXiv数学优化论文，研究分散式在线优化与连续次模最大化，与超节点、AI Rack、机柜级AI基础设施、供电散热高速互连及量产落地等主题完全无关，仅因英文单词Oracle命中搜索词，属于明显无关内容。"
AI质检模型: "zj-deepseek-v4-flash"
AI质检时间: "2026-09-02T03:57:08+08:00"
AI主题相关性: 0
AI来源权威性: 3
AI新颖性: 2
AI技术细节: 1
AI商业部署信号: 0
AI完整性: 2
AI摘要: "论文提出Dec-BFTRL算法，用于分离访问下的分布式在线上线性化优化以及连续DR-submodular最大化。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-01T23:06:14.166Z"
采集批次: "2026年9月1日23点56分52秒"
采集批次ID: "20260901-235652-761"
去重键: "https://arxiv.org/abs/2608.30271"
---

## Mathematics > Optimization and Control

## Title:Dec-BFTRL: Squre-Root Regret for Decentralized Online Upper-Linearizable Optimization under Separation Access with Application to Continuous Submodular Maximization

Authors:[Yiyang Lu](https://arxiv.org/search/math?searchtype=author&query=Lu,+Y), [Mohammad Pedramfar](https://arxiv.org/search/math?searchtype=author&query=Pedramfar,+M), [Vaneet Aggarwal](https://arxiv.org/search/math?searchtype=author&query=Aggarwal,+V)

[View PDF](https://arxiv.org/pdf/2608.30271) [HTML (experimental)](https://arxiv.org/html/2608.30271v1)

> Abstract: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 four wrapper instantiations covering three DR-submodular maximization problems.

| Subjects: | Optimization and Control (math.OC); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| --- | --- |
| Cite as: | [arXiv:2608.30271](https://arxiv.org/abs/2608.30271) \[math.OC\] |
|  | (or [arXiv:2608.30271v1](https://arxiv.org/abs/2608.30271v1) \[math.OC\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.30271](https://doi.org/10.48550/arXiv.2608.30271) |

## Submission history

From: Yiyang Lu \[[view email](https://arxiv.org/show-email/4b7954d7/2608.30271)\]  
**\[v1\]** Mon, 31 Aug 2026 05:29:54 UTC (40 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.30271) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))