Fast decentralized non-convex finite-sum optimization with recursive variance reduction

This paper considers decentralized minimization of $N:=nm$ smooth non-convex cost functions equally divided over a directed network of $n$ nodes. Specifically, we describe a stochastic first-order gradient method, called GT-SARAH, that employs a SARAH-type variance reduction technique and gradient tracking (GT) to address the stochastic and decentralized nature of the problem. We show that GT-SARAH, with appropriate algorithmic parameters, finds an $epsilon$-accurate first-order stationary point with $Obig(maxbig{N^{frac{1}{2}},n(1-lambda)^{-2},n^{frac{2}{3}}m^{frac{1}{3}}(1-lambda)^{-1}big}Lepsilon^{-2}big)$ gradient complexity, where ${(1-lambda)in(0,1]}$ is the spectral gap of the network weight matrix and $L$ is the smoothness parameter of the cost functions. This gradient complexity outperforms that of the existing decentralized stochastic gradient methods. In particular, in a big-data regime such that ${n = O(N^{frac{1}{2}}(1-lambda)^{3})}$, this gradient complexity furthers reduces to ${O(N^{frac{1}{2}}Lepsilon^{-2})}$, independent of the network topology, and matches that of the centralized near-optimal variance-reduced methods. Moreover, in this regime GT-SARAH achieves a non-asymptotic linear speedup, in that, the total number of gradient computations at each node is reduced by a factor of $1/n$ compared to the centralized near-optimal algorithms that perform all gradient computations at a single node. To the best of our knowledge, GT-SARAH is the first algorithm that achieves this property. In addition, we show that appropriate choices of local minibatch size balance the trade-offs between the gradient and communication complexity of GT-SARAH. Over infinite time horizon, we establish that all nodes in GT-SARAH asymptotically achieve consensus and converge to a first-order stationary point in the almost sure and mean-squared sense.