On Biased Compression for Distributed Learning
This work addresses communication efficiency in distributed learning, offering theoretical insights and new compressors that could improve performance in practical applications, though it is incremental in advancing existing compression techniques.
The paper tackles the problem of communication bottlenecks in distributed learning by studying biased compression operators, showing for the first time that they can achieve linear convergence rates in both single-node and distributed settings, with a proven ergodic rate of O(δL exp[-μK/δL] + (C + δD)/Kμ).
In the last few years, various communication compression techniques have emerged as an indispensable tool helping to alleviate the communication bottleneck in distributed learning. However, despite the fact biased compressors often show superior performance in practice when compared to the much more studied and understood unbiased compressors, very little is known about them. In this work we study three classes of biased compression operators, two of which are new, and their performance when applied to (stochastic) gradient descent and distributed (stochastic) gradient descent. We show for the first time that biased compressors can lead to linear convergence rates both in the single node and distributed settings. We prove that distributed compressed SGD method, employed with error feedback mechanism, enjoys the ergodic rate $O\left( δL \exp \left[-\frac{μK}{δL}\right] + \frac{(C + δD)}{Kμ}\right)$, where $δ\ge 1$ is a compression parameter which grows when more compression is applied, $L$ and $μ$ are the smoothness and strong convexity constants, $C$ captures stochastic gradient noise ($C=0$ if full gradients are computed on each node) and $D$ captures the variance of the gradients at the optimum ($D=0$ for over-parameterized models). Further, via a theoretical study of several synthetic and empirical distributions of communicated gradients, we shed light on why and by how much biased compressors outperform their unbiased variants. Finally, we propose several new biased compressors with promising theoretical guarantees and practical performance.