456048afb7253926e1fbb7486e699180-AuthorFeedback.pdf

The approximation errors are shown in Fig. (a)-(d). When the positive density region takes a large fraction of the20 entire domain, our method is very competitive (see Q3). When it is only inside a small part of a large domain, our21 method maynot be better than sequential adaptivesampling. Our subgroup rank-1 lattice performs good and robust. It does not26 haveany hyperparameter and is very convenient and cheap for points set construction.