Giuseppe Carleo

I don't know, I tend to like more the idea of being mathematically consistent and adopt the first solution proposed by Clemens

I think it would be good if we could make this work in a way that each rank has a fixed number of replicas, rather than a number of chains...

the first one, namely you can mix up and down spin orbitals, that's what unrestricted HF is (and yes, it makes sense)

ok so I think the naming is actually fine, what I wanted/referred to is the **generalized hartree fock**, and could be implemented quite easily adding a further flag to Slater2nd...

Yes I would use a constant number of discarded (not a fraction) but I wouldn't be in favor of having backend dependent defaults,this might be dangerous and give different results...

yes I agree this naming can be confusing, why not keeping only one MetropolisExchange that works for everything?

yes that might not be ideal uhm, I don't know... maybe let's just find an altenartive name then. I suggest "MetropolisHop"

essentially because you are doing a hopping of a particle on a lattice, which distinguishes enough with continuos space I guess

it would be ideal if `MetropolisHop` worked also for bosons, by the way!

I think this sampler should just take a random particle (whatever its spin) and move it in an empty site, so this is generalizable to all kind of particles on...