22
2

Cost-Efficient Distributed Learning via Combinatorial Multi-Armed Bandits

Abstract

We consider the distributed SGD problem, where a main node distributes gradient calculations among nn workers. By assigning tasks to all the workers and waiting only for the kk fastest ones, the main node can trade-off the algorithm's error with its runtime by gradually increasing kk as the algorithm evolves. However, this strategy, referred to as adaptive kk-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 kk workers, and gradually increases kk. 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 kk-sync, our scheme achieves significantly lower errors with the same computational efforts and less downlink communication while being inferior in terms of speed.

View on arXiv
Comments on this paper