arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2603.09168cs.LGcs.DSstat.ML
收藏

分布式专家问题的更优界限

Better Bounds for the Distributed Experts Problem

Carnegie Mellon University(卡内基梅隆大学);Texas A&M University(德克萨斯大学)

David P. Woodruff, Samson Zhou

AI总结:

本文提出了一种改进的分布式专家问题协议,通过更优的通信效率和悔恨值界限,提升了性能表现。

摘要:

在本文中,我们研究了分布式专家问题,其中n个专家被分布在s个服务器上,持续T个时间步。每个专家在每个时间t的损失是向量的ℓ_𝑝范数,该向量由每个服务器在时间t时该专家的损失组成。目标是通过最小化悔恨值R来实现最优性能,即分布式协议的损失与最佳专家损失的差值,按所有T次时间步平均,同时使用最小的通信量。我们提出了一种协议,其悔恨值约为R ≳ 1/(√T·poly log(nsT)),通信量为O(n/R² + s/R²)·max(s^{1-2/p},1)·poly log(nsT)位,这优于之前的工作。

英文摘要:

In this paper, we study the distributed experts problem, where $n$ experts are distributed across $s$ servers for $T$ timesteps. The loss of each expert at each time $t$ is the $\ell_p$ norm of the vector that consists of the losses of the expert at each of the $s$ servers at time $t$. The goal is to minimize the regret $R$, i.e., the loss of the distributed protocol compared to the loss of the best expert, amortized over the all $T$ times, while using the minimum amount of communication. We give a protocol that achieves regret roughly $R\gtrsim\frac{1}{\sqrt{T}\cdot\text{poly}\log(nsT)}$, using $\mathcal{O}\left(\frac{n}{R^2}+\frac{s}{R^2}\right)\cdot\max(s^{1-2/p},1)\cdot\text{poly}\log(nsT)$ bits of communication, which improves on previous work.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑