At current we use sampling with replacement, but a coin flip per node (c.f. percolation process) would be better I think.
I think that I imported the concept of bootstrap sampling a little too literally from phylogenetics, whereby alignment sites are sampled with replacement.
Re: bootstrap sampling of fixed terminals.
At current we use sampling with replacement, but a coin flip per node (c.f. percolation process) would be better I think.
I think that I imported the concept of bootstrap sampling a little too literally from phylogenetics, whereby alignment sites are sampled with replacement.