This paper presents a new generalization error analysis for Decentralized Stochastic Gradient Descent (D-SGD) based on algorithmic stability. The obtained results overhaul a series of recent works that suggested an increased instability due to decentralization and a detrimental impact of poorly-connected communication graphs on generalization. On the contrary, we show, for convex, strongly convex and non-convex functions, that D-SGD can always recover generalization bounds analogous to those of classical SGD, suggesting that the choice of graph does not matter. We then argue that this result is coming from a worst-case analysis, and we provide a refined data-dependent generalization bound for general convex functions. This new bound reveals that the choice of graph can in fact improve the worst-case bound in certain regimes, and that surprisingly, a poorly-connected graph can even be beneficial.

翻译：本文基于算法稳定性理论，对去中心化随机梯度下降（D-SGD）算法提出了新的泛化误差分析。所得结果推翻了近期一系列认为去中心化会导致不稳定性增加、低连通性通信图会损害泛化性能的研究结论。相反，我们证明在凸函数、强凸函数和非凸函数情形下，D-SGD始终能够恢复与经典SGD相当的泛化界，这表明图的选择无关紧要。随后我们论证该结果源于最坏情况分析，并针对一般凸函数提出了改进的数据依赖型泛化界。这一新界限揭示：图的选择实际上可以在特定情况下改进最坏情况界，且令人惊讶的是，低连通性图甚至可能具有优势。