[feature request] Partially labeled SBM in graph-tool
As far as I understand, SBM are a class of techniques which in the machine-learning community would be called "unsupervised models". Given the graph we are going to estimate the labels of nodes, to find the assignation which maximizes the likelihood of the model.
However, it would be nice to include semi-supervised models too in terms of SBM. Given that I already know some of the labels, how can I estimate the others? This is what the old label-propagation does, but without all the nice theoretical machinery of SBM which makes the latter a much better tool. I've found this page interesting.
Is there already in graph-tool any way to specify known and unknown label in the model as, for example, node covariates to be considered fixed?