Archive January 2022

Evidence on tap: we introduce "floZ", a new method to compute the Bayesian evidence from pre-existing posterior samples based on normalizing flows.

We tested it in up to 15D, finding good agreement with nested sampling 😃

Check it out:

https://arxiv.org/abs/2404.12294

Fellow academics: if you use "significantly" a lot, you are beginning to sound like ChatGPT! 😎
👏 lead author @GengMingmeng for teasing out the impact of LLM's from 1 million preprint abstracts.
tldr: computer scientists appear to use it a lot more than mathematicians.

Where am I?