search

joshspeagle / dynesty

Dynamic Nested Sampling package for computing Bayesian posteriors and evidences
https://dynesty.readthedocs.io/
MIT License
347 stars 76 forks source link

Feature request: add support for blob data #375

Closed cdcapano closed 2 years ago

cdcapano commented 2 years ago

It would be great if dynesty could support storing "blob" data. This is extra metadata about sample points; see, for example, emcee's documentation on blob data here.

We (PyCBC) have extra metadata about samples that we like to store for gravitational-wave parameter estimation. Currently, when using dynesty, we need to run followup scripts that recalculate the blob data for every sample point and add it to our results files. This is a bit cumbersome and unnecessarily costly. The information is generated by our models when their loglikelihood function is called, we just have no way of passing it to dynesty to store it in memory for retrieval after the run is over.

segasai commented 2 years ago

I agree -- this is essentially the same issue as #368

cdcapano commented 2 years ago

Ah, sorry, didn't see that.

segasai commented 2 years ago

I'll close this issue for the time being (it'll be still linked from #368). But the blob interface in emcee is a good pointer for possible interface to try to replicate here