arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2202.10862cs.DC
收藏

基于集群模型的异步全去中心化SGD

Asynchronous Fully-Decentralized SGD in the Cluster-Based Model

Hagit Attiya, Noa Schiller

AI总结:

本文面向结合消息传递与共享内存的集群模型,提出可容错的异步全去中心化SGD算法,针对强凸函数实现最大化分布式加速,对任意函数通过定制多维近似一致性算法达到接近串行的收敛速率,并证明多数集群含非故障进程的必要性。

摘要:

本文提出了具备容错能力的异步随机梯度下降(Stochastic Gradient Descent,SGD)算法。SGD作为优化与学习算法的核心组件,被广泛用于逼近代价函数$Q$的最小值。我们的算法面向结合了消息传递与共享内存通信层的集群模型(cluster-based model)设计。进程可能发生崩溃故障,每个集群内部的算法是无等待的,仅使用读和写操作。\n 对于强凸函数$Q$,我们的算法可容忍任意数量的故障,且其收敛速率相较于串行SGD的最优收敛速率实现了最大化的分布式加速。\n 对于任意函数,收敛速率存在一个额外项,该项取决于同一迭代中各参数之间的最大差值(这一结论在关于$Q$的标准假设下成立)。在这种情况下,算法能达到与串行SGD相同的收敛速率,仅相差一个对数因子。这一结果通过在每次迭代中使用专为集群模型定制的多维近似一致性算法实现。\n 针对任意函数的算法要求至少半数以上的集群包含至少一个非故障进程。我们证明,在优化某些非凸函数时,该条件是必要的。

英文摘要:

This paper presents fault-tolerant asynchronous Stochastic Gradient Descent (SGD) algorithms. SGD is widely used for approximating the minimum of a cost function $Q$, as a core part of optimization and learning algorithms. Our algorithms are designed for the cluster-based model, which combines message-passing and shared-memory communication layers. Processes may fail by crashing, and the algorithm inside each cluster is wait-free, using only reads and writes. For a strongly convex function $Q$, our algorithm tolerates any number of failures, and provides convergence rate that yields the maximal distributed acceleration over the optimal convergence rate of sequential SGD. For arbitrary functions, the convergence rate has an additional term that depends on the maximal difference between the parameters at the same iteration. (This holds under standard assumptions on $Q$.) In this case, the algorithm obtains the same convergence rate as sequential SGD, up to a logarithmic factor. This is achieved by using, at each iteration, a multidimensional approximate agreement algorithm, tailored for the cluster-based model. The algorithm for arbitrary functions requires that at least a majority of the clusters contain at least one nonfaulty process. We prove that this condition is necessary when optimizing some non-convex functions.

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。

↑