A Momentum-Guided Frank-Wolfe Algorithm

With the well-documented popularity of Frank Wolfe (FW) algorithms in machine learning tasks, present paper establishes links between FW subproblems and notion momentum emerging accelerated gradient methods (AGMs). On one hand, these reveal why is unlikely to be effective for FW-type on general problems. other it established that accelerates a class signal processing applications. Specifically, proved variant FW, here termed (AFW), converges with faster rate O (\frac1k 2 ) such family problems, despite same (\frac1k) cases. Distinct from existing fast convergent variants, rates rely parameter-free step sizes. Numerical experiments benchmarked tasks corroborate theoretical findings.