Beware local optima.
This approach is essentially a maximum-likelihood sample, and like normal sampling this can have issues where there are very good regions of state space “down in the tree”, but no good way to get there.
Beware local optima.
This approach is essentially a maximum-likelihood sample, and like normal sampling this can have issues where there are very good regions of state space “down in the tree”, but no good way to get there.