they pointed out the importance of the problem and of our analysis given the increasing popularity of computational

We would like to thank the reviewers for their efforts on reading and evaluating our paper . In what follows, we provide specific responses to each reviewer. We thank R1 for a very detailed evaluation and truly helpful feedback on our empirical analysis. The hyperparameter tuning [...] and initialization steps [...] are not explained clearly enough We will move necessary Can't we apply the same strategy in the Frank-W olfe algorithm of SRW [Alg.2, 64]: Y es, we can apply the same Indeed, we made good use of the open source code they have provided to inspire our implementation. Is it possible to extend this result to PRW?