Joshua G Albert

Can we have this?

Having suffered from this problem for some time, I'm happy to find a thread that fights to get this feature. (It shouldn't even be considered a feature as far as...

My pleasure. I don't have access to that setting from enduser state. We're integrating ultranest into bilby, and then running on a set of problems (https://github.com/Joshuaalbert/jaxns-cosmology/blob/main/jaxns_cosmology/models/__init__.py). The integration is here:...

I'll try that out.

I see more likelihood evaluations are used when `max_num_improvement_loops=-1`. ``` { "model_name": "CMB", "params": { "min_num_live_points": 150, "max_ncalls": 50000000, "max_num_improvement_loops": -1 }, "likelihood_evals": 156466 } ``` ``` { "model_name": "CMB",...

It's not an evidence comparison that is being done, but rather a posterior inference test. Also, we're assuming the default stopping conditions for each sampler, as they are all different.

AffineLayer grid search is almost over, but I can say this already: | Model | AffineLayer | MaxPrincipleGapAffineLayer | |---|---|---| | CMB | Works | Works | | EggBox10D |...

It's the same to current version with AffineLayer. Though I also searched over Reactive and Static samplers, and the runs that I saved used the fewest number of live points...

I willOn Mar 7, 2024 13:55, Johannes Buchner ***@***.***> wrote: Please be sure to label the results in your comparison as "MLFriends" or "UltraNest/MLFriends", as that is the algorithm used...

6 out of 8 tests fail ``` ========================= 6 failed, 2 passed in 1.72s ========================== FAILED [ 12%] debug/error.py:8 (test_truncated_normal[10-0.00-0.01]) Array([ -inf, 0.9767743 , 0.97950405, 0.9812364 , 0.98253983, 0.98360026, 0.9845029...