Optimal algorithms for smooth and strongly convex distributed optimization in networks
This work provides foundational algorithms for efficient distributed optimization, impacting fields like machine learning and data analysis by enabling faster convergence in networked systems.
The paper determines optimal convergence rates for distributed optimization in networks, showing that distributing Nesterov's accelerated gradient descent is optimal for centralized settings and introducing the multi-step dual accelerated (MSDA) method as the first optimal algorithm for decentralized gossip-based settings, achieving precision in time O(√κ_l(1+τ/√γ)ln(1/ε)).
In this paper, we determine the optimal convergence rates for strongly convex and smooth distributed optimization in two settings: centralized and decentralized communications over a network. For centralized (i.e. master/slave) algorithms, we show that distributing Nesterov's accelerated gradient descent is optimal and achieves a precision $\varepsilon > 0$ in time $O(\sqrt{κ_g}(1+Δτ)\ln(1/\varepsilon))$, where $κ_g$ is the condition number of the (global) function to optimize, $Δ$ is the diameter of the network, and $τ$ (resp. $1$) is the time needed to communicate values between two neighbors (resp. perform local computations). For decentralized algorithms based on gossip, we provide the first optimal algorithm, called the multi-step dual accelerated (MSDA) method, that achieves a precision $\varepsilon > 0$ in time $O(\sqrt{κ_l}(1+\fracτ{\sqrtγ})\ln(1/\varepsilon))$, where $κ_l$ is the condition number of the local functions and $γ$ is the (normalized) eigengap of the gossip matrix used for communication between nodes. We then verify the efficiency of MSDA against state-of-the-art methods for two problems: least-squares regression and classification by logistic regression.