Wrong potential resulting from reordering the states of its parent

[Jorge Pérez]

In the attached network, if we reorder the states of the variable “Intervencion”, the potential of “C: Dispositivo” is no longer valid. More exactly, although the cost of the ICB should be 21000, when the parent states are reordered, this value is assigned to the cost of the ICU.

[Manuel Luque]

The current implementation of the methods for reordering the states of a variable covers the case in which the potential is a TablePotential. Based on such implementation, with minimal changes, I am going to extend it to ExactDistrPotential. However, the current implementation does not cover the case of TreeADDs, for example. We could create an issue with this proposal of enhancement (or at least to discuss in next meeting).

[Manuel Luque]

This commit fixes this issue.