Stochastic Primal-Dual Coordinate Method for Regularized Empirical Risk Minimization
This provides an efficient optimization method for machine learning practitioners dealing with large-scale empirical risk minimization problems, though it appears incremental relative to existing primal-dual approaches.
The authors tackled the problem of regularized empirical risk minimization for linear predictors by proposing a stochastic primal-dual coordinate (SPDC) method that reformulates it as a saddle point problem and achieves accelerated convergence. They showed theoretically and empirically that SPDC performs comparably or better than state-of-the-art methods.
We consider a generic convex optimization problem associated with regularized empirical risk minimization of linear predictors. The problem structure allows us to reformulate it as a convex-concave saddle point problem. We propose a stochastic primal-dual coordinate (SPDC) method, which alternates between maximizing over a randomly chosen dual variable and minimizing over the primal variable. An extrapolation step on the primal variable is performed to obtain accelerated convergence rate. We also develop a mini-batch version of the SPDC method which facilitates parallel computing, and an extension with weighted sampling probabilities on the dual variables, which has a better complexity than uniform sampling on unnormalized data. Both theoretically and empirically, we show that the SPDC method has comparable or better performance than several state-of-the-art optimization methods.