Hello ! I'm a physicist working on Many-Body Quantum Problems using ML. I'm currently trying to use Darkon to quantify which points of my dataset might be kept and which ther are Harmfull.
Very basically i have quite of a deep network implemented via Keras API which performs a regression task
Is Darkon works with MSE loss ? It'sappear that ther is a sign problem occuring when trying to discriminate between Harmfull/Helpfull samples