Ultranest version 4

[JohannesBuchner]

Dear users,

you recently interacted with UltraNest on github. I just pushed a new version, 4.1, to conda and pypi. You can find the changes and merged pull requests (huge thanks!) in HISTORY.rst

I would really appreciate if you downloaded it and tested it in your workflows. In particular, the default algorithm MLFriends is now run with a slightly different way of computing a local distance metric (layer_class=MaxPrincipleGapAffineLayer instead of AffineLayer), and I wonder whether that makes fits slightly more efficient (or not!).

Step samplers have been gifted a new diagnostic (relative jump distance), which I am currently writing up as a paper.

@comane @odstrcilt @newmandb @HoisW @OGdodge @gregorydavidmartinez @adipol-ph @facero @gregorydavidmartinez @lwelzel @jpl-jengelke @ahnitz @PieterVuylsteke @ikhebgeenaccount @jacopok

If you find any problems, please open a new issue.

[facero]

Hi Thanks. Will test it soon in my workflow. Just looked at the History.rst and it seems there is a typo in the dates of the release (2023 => 2024). Le dim. 18 févr. 2024, 12:38, Johannes Buchner ***@***.***> a écrit :

…

Dear users, you recently interacted with UltraNest on github. I just pushed a new version, 4.1, to conda and pypi. You can find the changes and merged pull requests (huge thanks!) in HISTORY.rst <https://github.com/JohannesBuchner/UltraNest/blob/master/HISTORY.rst> I would really appreciate if you downloaded it and tested it in your workflows. In particular, the default algorithm MLFriends is now run with a slightly different way of computing a local distance metric (layer_class=MaxPrincipleGapAffineLayer instead of AffineLayer), and I wonder whether that makes fits slightly more efficient (or not!). Step samplers have been gifted a new diagnostic (relative jump distance), which I am currently writing up as a paper. @comane <https://github.com/comane> @odstrcilt <https://github.com/odstrcilt> @newmandb <https://github.com/newmandb> @HoisW <https://github.com/HoisW> @OGdodge <https://github.com/OGdodge> @gregorydavidmartinez <https://github.com/gregorydavidmartinez> @adipol-ph <https://github.com/adipol-ph> @facero <https://github.com/facero> @gregorydavidmartinez <https://github.com/gregorydavidmartinez> @lwelzel <https://github.com/lwelzel> @jpl-jengelke <https://github.com/jpl-jengelke> @ahnitz <https://github.com/ahnitz> @PieterVuylsteke <https://github.com/PieterVuylsteke> @ikhebgeenaccount <https://github.com/ikhebgeenaccount> @jacopok <https://github.com/jacopok> If you find any problems, please open a new issue. — Reply to this email directly, view it on GitHub <#121>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACCAQYUPSCXBS343MVM23ADYUHY2VAVCNFSM6AAAAABDOECGU6VHI2DSMVQWIX3LMV43ASLTON2WKOZSGE2DAOJWHAYTSOA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[JohannesBuchner]

Very good, thank you!

[jacopok]

Hi, I made a quick test with some low-dimensional, simple fits I was running (5 parameters), and I am observing no major difference in the run time on my machine from version 3.6 to version 4.1.
I did not investigate this thoroughly - the fits took on the order of a few seconds, and the number of seconds was the same after the update.

[JohannesBuchner]

Thanks for the report @jacopok!