2 SCENT
Detailed descriptions are given [Butchart et al., 1997]
so here only a summary of the model is given.
Learning in SCENT takes place by a simple recursive
search through the tree, with the winner in each subtree,
the node whose weight vector is nearest in Euclidean
space, moving towards the current input vector. As in
other neural clusterers the movement is mediated by a
learning rate.