Closed agoscinski closed 1 year ago
I think this is an abstraction that is not important for people who are interested the DCH. If we need such an abstraction we can revive this issues, but I don't see the necessity. This issue was just to keep track of some ideas in the conception phase of the DCH.
discussed with @victorprincipe a possible redesign of GCH in some future to have additional functionalities and more flexibility. At the moment it is just used scorer. It has hard to make a estimator out of the GCH, because to be consistent with the score function we would need to predict X_HD, but in the current fit function X_HD is part of the X input, so it would be an inconsistent in fit and predict function if we predict X_HD.
So we thought about separating the current DCH directional convex hull and the interpolator part into two separated classes and bringing them then together
I feel like the last part cannot be abstracted and has to be just a recipe in an example notebook because it is such a specific use case.