The world is being quietly rearranged by people who write very long documents.


The title they went with Curved representational Bregman divergences and their applications Noisy translates that to

Mathematical framework for measuring similarity on curved spaces


Researchers extended a mathematical tool called Bregman divergence — which measures distance between points — to work on curved surfaces and restricted subspaces, not just flat geometric spaces. This makes it possible to calculate meaningful distances and averages for data that naturally lives on curved structures, like normalized vectors or probability distributions.
This is pure mathematics with no demonstrated real-world deployment, economic impact, or evidence it changes what's actually computable in practice at scale.

If you insist
Read the original →