A generalized Frank–Wolfe method with “dual averaging” for strongly convex composite optimization

Abstract We propose a simple variant of the generalized Frank–Wolfe method for solving strongly convex composite optimization problems, by introducing an additional averaging step on dual variables. show that in this variant, one can choose constant step-size and obtain linear convergence rate duality gaps. By leveraging analysis we then analyze local logistic fictitious play algorithm, which is well-established game theory but lacks any form guarantees. that, with high probability, algorithm converges locally at O (1/ t ), terms certain expected gap.