joshspeagle / dynesty

Dynamic Nested Sampling package for computing Bayesian posteriors and evidences
https://dynesty.readthedocs.io/
MIT License
357 stars 77 forks source link

vol underflow in high dimensions #91

Closed joshspeagle closed 6 years ago

joshspeagle commented 6 years ago

For ndim large, vol_prefactor, linalg.det, and scale_to_vol can under/overflow. Switch to log-based arithmetic to avoid this behavior for bounding ellipsoids. (Other bounding schemes shouldn't be used when ndim is large anyway.)