And may I ask another question? In the paper, it mentioned "by making
beta goes to infinity and repeated many times, which yields a reliable
estimate of the maximum". May I double-check what "repeated many times"
refers to? Does it refer to the number of sweeps or refer to the whole
algorithm?
The whole algorithm.
I also noticed there is a warning of "multilevel_mcmc_sweep"
in NestedBlockState: "This function performs niter sweeps at each
hierarchical level once. This means that in order for the chain to
equilibrate, we need to call this function several times, i.e. it is not
enough to call it once with a large value of niter." I found that the
high-level function "minimize_nested_blockmodel_dl" seems already done
that. But I am a little bit confused, if possible, could it be more
specific? Thank you so much for your help!
This warning is not applicable if minimize_nested_blockmodel_dl() is
being used, only if you are attempting to sample from the posterior
distribution (and not finding its maximum).
Best,
Tiago