Explore chapters and articles related to this topic
Stochastic approximation versus sample average approximation for Wasserstein barycenters
Published in Optimization Methods and Software, 2022
Darina Dvinskikh
Using Dual Extrapolation method for the WP problem [25] instead of Mirror Prox, we can omit in the last row of Table 2. One of the main advantages of the SAA approach is the possibility to perform it in a decentralized manner in contrast to the SA approach, which cannot be executed in a decentralized manner or even in distributed or parallel fashion for non-smooth objective [36]. This is the case of the Wasserstein barycenter problem, indeed, the objective is Lipschitz continuous but not Lipschitzsmooth.