Johannes Buchner

Thanks for the report @jacopok!

There are some example light curves in https://github.com/JohannesBuchner/bexvar/tree/main/examples The format is documented in https://erosita.mpe.mpg.de/edr/DataAnalysis/srctool_doc.html

Possibly, but that depends on many factors.

Thank you for this very useful feedback. I guess you are running without a step sampler. The main change that occurred there is the introduction of MaxPrincipleGapAffineLayer. Can you please...

Can you do the following? ``` sampler = ReactiveNestedSampler(...) sampler.transform_layer_class = ultranest.mlfriends.AffineLayer ```

Well, you are supposed to get a refined ln(Z) estimate and posterior distribution too with higher evals.

For a fair comparison across samplers you should compare cost for the same quality, which may be ln(Z) scatter around the true value, or distribution of the posterior z-score at...

I think what you see is expected: you should see higher ESS for higher computational cost runs, and a posterior comparison across samplers should consider this. While `max_num_improvement_loops!=0` with a...

Thank you for sharing these results. Just for clarification, are the results for 3.5 better (more often completed?)

Please be sure to label the results in your comparison as "MLFriends" or "UltraNest/MLFriends", as that is the algorithm used here. UltraNest can use several algorithms, including slice sampling, which...