search

joshspeagle / dynesty

Dynamic Nested Sampling package for computing Bayesian posteriors and evidences
https://dynesty.readthedocs.io/
MIT License
358 stars 78 forks source link

random walk sampling stuck - scale hitting 0.0 #85

Closed villrv closed 6 years ago

villrv commented 6 years ago

Have been running dynesty with mosfit and receive the following error before convergence:

"RuntimeError: The random walk sampling is stuck! Some useful output quantities:...."

due to the scale hitting 0.0. Is there a way to avoid this issue?

joshspeagle commented 6 years ago

This is the first time I've seen this issue arise with the random walk sampler. Can you send me the exact outputs from this or the code/seed you used that produced this result? And is this with the most recent version of dynesty (v0.9.1)?

joshspeagle commented 6 years ago

@villrv and I weren't able to consistently reproduce this issue under the same conditions, so I'm closing this for now. We can come back to it if it continues to be a problem.

joshspeagle commented 6 years ago

After some debugging with help from @guillochon, it looks like this is a problem where the likelihood "slams" against the bounds of the unit cube in >=1 dimension of the fit. I'll try and work on a fix to deal with this issue ASAP.

joshspeagle commented 6 years ago

This seems to have been mostly resolved, so I'm re-closing this issue for now barring something catastrophic.