From Newsgroup: comp.arch
I just stumbled across this article on probabilistic rounding:
https://pmc.ncbi.nlm.nih.gov/articles/PMC8905452/ (not behind
paywall, fortunately).
Quite interesting. They propose (quoting from the text)
"[...] round x to the next larger or smaller number y ree F
with probability 1 minus the distance between x and y divided by
rieEYaNrie reArieEYaNrii."
plus give an algorithm.
Their ODE example is a bit suspect - nobody but the truly clueless
use the forward Euler method for calculating the solution to an
ordinary differential equation. It would have been nice to see
at least fourth/fifth order Runge-Kutta, or a predictor/corrector
method. (And solving PDEs via Runge-Kutta in one direction and
finite differences in the other is also... well, some people
like to stretch their tools, in this case their ODE solvers).
Still, good reading.
--
This USENET posting was made without artificial intelligence,
artificial impertinence, artificial arrogance, artificial stupidity,
artificial flavorings or artificial colorants.
--- Synchronet 3.21a-Linux NewsLink 1.2