Title:Stochastic Variance-Reduced Forward-Reflected Methods for Root-Finding Problems

Abstract:We develop two novel stochastic variance-reduction methods to approximate a solution of root-finding problems applicable to both equations and inclusions. Our algorithms leverage a new combination of ideas from the forward-reflected-backward splitting method and a class of unbiased variance-reduction estimators. We construct two new stochastic estimators within this class, inspired by the well-established SVRG and SAGA estimators. These estimators differ significantly from existing approaches used for root-finding algorithms. By appropriately selecting parameters, both algorithms achieve the state-of-the-art oracle complexity of $\mathcal{O}(n + n^{2/3} \epsilon^{-2})$ for achieving an $\epsilon$-solution in terms of the operator residual norm, where $n$ represents the number of summands and $\epsilon$ signifies the desired accuracy. This complexity aligns with the best-known results in stochastic nonconvex optimization without enhancements. We test our algorithms on two numerical examples and compare them with existing methods. The results demonstrate promising improvements offered by the new methods compared to their competitors.