A short observation about per-displayed-tree-clvs

[lutteropp]

Naively, one would store number_of_network_nodes * number_of_displayed_trees CLV vectors. However, there are regions in the network that are exactly the same over multiple displayed trees. Those regions can be identified with the blob optimization trick.

---> One actually can save quite some memory and computations in there, if one cleverly reuses some CLVs/ shares them among multiple displayed trees. It will be tricky to make this work together with dynamical changes (network moves) to the network data structure.