Closed rory-smith closed 4 years ago
Yes, this is indeed a possible issue and one dynesty
deals with by (1) avoiding constructing the first ellipsoids until the points are reasonably far away from the edges and (2) automatically rejecting all out-of-bounds samples (except in cases where reflective
or periodic
bounds are enabled). The place where this occurs with the MultiEllipsoid
sampler is here.
Hi Josh,
I'd like to understand better how I can modify dynesty to avoid the issue described in the subject heading. I assume that the bounding ellipses can go out of the bounds of the unit cube which can cause samples within the ellipses to also be out of bounds. I'd be grateful if you could point me to the right place to look to see where uniform samples are projected onto a bounding ellipse so I can play around with this.
Best, Rory