arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2202.08302cs.ITcs.DCcs.LGmath.ITstat.ML
收藏

基于组合多臂Bandits的高效分布式学习

Cost-Efficient Distributed Learning via Combinatorial Multi-Armed Bandits

Technical University of Munich(慕尼黑工业大学);Imperial College London(帝国理工学院)

Maximilian Egger, Rawad Bitar, Antonia Wachter-Zeh, Deniz Gündüz

摘要:

我们考虑分布式SGD问题,其中一个主节点分布梯度计算之间的$n$工人。通过将任务分配给所有工作者并仅等待最快的$k$,主节点可以通过随着算法的发展逐渐增加$k$来权衡算法的错误与其运行时间。然而,这种称为自适应$k$ -sync的策略忽略了未使用的计算的成本以及将模型传达给揭示离散行为的工作者的成本。我们提出了一个具有成本效益的计划,只分配任务的$k$工人,并逐步增加$k$。我们介绍了使用一个组合的多臂强盗模型来学习哪些工人是最快的,同时分配梯度计算。假设工人的响应时间呈指数分布,通过不同的方法进行参数化,我们给出了经验和理论上的保证,我们的策略,即遗憾,学习工作进程的平均响应时间所花费的额外时间。此外,我们提出并分析了一个策略适用于一个大类的响应时间分布。自适应$k$同步相比,我们的计划实现了显着降低的错误与相同的计算工作量和更少的下行链路通信,而在速度方面是低劣的。

英文摘要:

We consider the distributed SGD problem, where a main node distributes gradient calculations among $n$ workers. By assigning tasks to all the workers and waiting only for the $k$ fastest ones, the main node can trade-off the algorithm's error with its runtime by gradually increasing $k$ as the algorithm evolves. However, this strategy, referred to as adaptive $k$-sync, neglects the cost of unused computations and of communicating models to workers that reveal a straggling behavior. We propose a cost-efficient scheme that assigns tasks only to $k$ workers, and gradually increases $k$. We introduce the use of a combinatorial multi-armed bandit model to learn which workers are the fastest while assigning gradient calculations. Assuming workers with exponentially distributed response times parameterized by different means, we give empirical and theoretical guarantees on the regret of our strategy, i.e., the extra time spent to learn the mean response times of the workers. Furthermore, we propose and analyze a strategy applicable to a large class of response time distributions. Compared to adaptive $k$-sync, our scheme achieves significantly lower errors with the same computational efforts and less downlink communication while being inferior in terms of speed.

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。

↑