Giuseppe Carleo

I would say that more than redding things out on the acceptance, it would make more sense to red things out for high correlation times? ALPS had some coloring like...

Yes, I would add - ~Remove Legacy~ #773

I think that the clean way to achieve this is to extend MetropolisLocal in a way that it generates at random one state out of the local hilbert space spanned...

no, if you flip all spins it's not ergodic, you have to generate one of the 2^n possible local states

including those where you flip only one spin, so in a sense that's just a generalization of the case n=1 which is contained as a special case

it doesn't seem too hard actually, you need to generate the states at the beginning, calling Hilbert once, and then every time you just draw an integer in [0,local_size(n)-1]...

Yes, it'd be great to have this tutorial in v 2.0!

Nice

> We could have 5 different operators KineticOperator implementing the different methods? Wait, does that involve using multiple dispatch?!?! :) I vote for Jannes' proposal to create a `method` kwarg

if there is no performance hit, I would merge this