Releases · nicolaipalm/paref

This release includes major changes. The main changes are:

each algorithm now automatically determines optimal hyperparameters; this includes the pre-scaling of the training data for the gpr and the optimal "epsilon" in all Pareto reflections
storing the evaluations is now automatically done by the blackbox function interface
there is now an express version which wraps the most common MOO algorithms of Paref
there is now an info module providing information about the pareto front and the training process which guides the further optimization
you can now access the best fitting Pareto points corresponding to some Pareto reflection within the interface
saving and loading is now included within the blackbox function interface
lhc sampling is now excluded by the moo algorithms; this is because it was misleading when specifying the max number of evaluations

Provide feedback