Giuseppe Carleo

Ah ok but then I don't like the solution proposed, I'd rather implement this in the acceptance!

Just put an if in the acceptance, if the configurations are identical you claim it as rejected. A much better solution would be to propose to exchange only configurations that...

If you propose to exchange only different local states then you need to be careful with the actual Hastings correction, but it's a much cleaner solution in my view

I think it is wrong to claim that equal configurations are rejected, so I am against hacking the sampler.... I would instead add a flag that allows one to restrict...

What you need for the move to be efficient is to 1. Identify all sites that have different occupation numbers, say there are K of those 2. Pick a random...

The Hastings correction would be log(K) indeed, so if K is zero you don't accept, that's a clean solution

thanks Zakari I think we should definitely include the last code you wrote as a comment (maybe do a separate PR or within this one?) enabling the option to have...

> I am not sure I see what you mean, I indeed compare sigmas on both sides of the != operator: σ[..., rule.clusters[:, 0]] != σ[..., rule.clusters[:, 1]]. yeah sorry,...

also, it's likely this is more expensive than the simpler exchange rule, but I would expect this is certainly not the bottleneck, and even on small systems (i.e. the canonical...

I think this is a bug and should be corrected anyways?