We consider decentralized stochastic optimization with the objective function (e.g. data samples for machine learning task) being distributed over n machines that can only communicate to their neighbors on a fixed communication graph. To reduce the communication bottleneck, the nodes compress (e.g. quantize or sparsify) their model updates. We cover both unbiased and biased compression operators with quality denoted by ω ≤ 1 (ω = 1 meaning no compression). We (i) propose a novel gossip-based stochastic gradient descent algorithm, Choco-SGD, that converges at rate O 1/(nT ) + 1/(T δ 2 ω) 2 for strongly convex objectives, where T denotes the number of iterations and δ the eigengap of the connectivity matrix. Despite compression quality and network connectivity affecting the higher order terms, the first term in the rate, O(1/(nT )), is the same as for the centralized baseline with exact communication. We (ii) present a novel gossip algorithm, Choco-Gossip, for the average consensus problem that converges in time O(1/(δ 2 ω) log(1/ )) for accuracy > 0. This is (up to our knowledge) the first gossip algorithm that supports arbitrary compressed messages for ω > 0 and still exhibits linear convergence. We (iii) show in experiments that both of our algorithms do outperform the respective state-of-the-art baselines and Choco-SGD can reduce communication by at least two orders of magnitudes.

Decentralized training of deep learning models is a key element for enabling data privacy and on-device learning over networks, as well as for efficient scaling to large compute clusters. As current approaches suffer from limited bandwidth of the network, we propose the use of communication compression in the decentralized training context. We show that Choco-SGD-recently introduced and analyzed for strongly-convex objectives only-converges under arbitrary high compression ratio on general non-convex functions at the rate O 1/ √ nT where T denotes the number of iterations and n the number of workers. The algorithm achieves linear speedup in the number of workers and supports higher compression than previous state-of-the art methods. We demonstrate the practical performance of the algorithm in two key scenarios: the training of deep learning models (i) over distributed user devices, connected by a social network and (ii) in a datacenter (outperforming all-reduce time-wise).

We consider decentralized stochastic variational inequalities where the problem data is distributed across many participating devices (heterogeneous, or non-IID data setting). We propose a novel method-based on stochastic extragradient-where participating devices can communicate over arbitrary, possibly time-varying network topologies. This covers both the fully decentralized optimization setting and the centralized topologies commonly used in Federated Learning. Our method further supports multiple local updates on the workers for reducing the communication frequency between workers. We theoretically analyze the proposed scheme in the strongly monotone, monotone and non-monotone setting. As a special case, our method and analysis apply in particular to decentralized stochastic min-max problems which are being studied with increased interest in Deep Learning. For example, the training objective of Generative Adversarial Networks (GANs) are typically saddle point problems and the decentralized training of GANs has been reported to be extremely challenging. While SOTA techniques rely on either repeated gossip rounds or proximal updates, we alleviate both of these requirements. Experimental results for decentralized GAN demonstrate the effectiveness of our proposed algorithm.

Decentralized optimization methods enable on-device training of machine learning models without a central coordinator. In many scenarios communication between devices is energy demanding and time consuming and forms the bottleneck of the entire system. We propose a new randomized first-order method which tackles the communication bottleneck by applying randomized compression operators to the communicated messages. By combining our scheme with a new variance reduction technique that progressively throughout the iterations reduces the adverse effect of the injected quantization noise, we obtain the first scheme that converges linearly on strongly convex decentralized problems while using compressed communication only. We prove that our method can solve the problems without any increase in the number of communications compared to the baseline which does not perform any communication compression while still allowing for a significant compression factor which depends on the conditioning of the problem and the topology of the network. Our key theoretical findings are supported by numerical experiments.

In decentralized machine learning, workers compute model updates on their local data. Because the workers only communicate with few neighbors without central coordination, these updates propagate progressively over the network. This paradigm enables distributed training on networks without all-to-all connectivity, helping to protect data privacy as well as to reduce the communication cost of distributed training in data centers. A key challenge, primarily in decentralized deep learning, remains the handling of differences between the workers' local data distributions. To tackle this challenge, we introduce the RelaySum mechanism for information propagation in decentralized learning. RelaySum uses spanning trees to distribute information exactly uniformly across all workers with finite delays depending on the distance between nodes. In contrast, the typical gossip averaging mechanism only distributes data uniformly asymptotically while using the same communication volume per step as RelaySum. We prove that RelaySGD, based on this mechanism, is independent of data heterogeneity and scales to many workers, enabling highly accurate decentralized deep learning on heterogeneous data. Our code is available at http://github.com/epfml/relaysgd.