Better ensemble.autocorrelation_time

evanberkowitz / supervillain

MCMC for modified Villain constructions

https://supervillain.readthedocs.io/

Other

0 stars 0 forks source link

Better ensemble.autocorrelation_time #94

Open evanberkowitz opened 9 months ago

evanberkowitz commented 9 months ago

In #79 I mentioned that it would be nice to be able to provide a cut before computing the autocorrelation time. This really ought to wait until after #80. To have a truly fully-featured autocorrelation_time method it would be great to be able to

[x] cut from the beginning for thermalization
[ ] specify means if known. For example, XWrapping has a known mean of 0 by symmetry; if an ensemble got trapped it might falsely look decorrelated.

evanberkowitz commented 1 week ago

Specifying the means is extremely annoying to accomplish, but might be doable by adding features to Observables.