Find possible loss functions for xyzt and benchmark these against each other

Created by: RasmusOrsoe

We need a xyzt regressor (but we dont have one!).

What we know

• Bjørn and Mads treated xyz and t independently. (See attached thesis from Bjørn, page 57 for loss function and consideration)

To do:

• Use Bjørn and Mads' loss function and train dynedge with it
• evaluate vertex regression performance w.r.t. retro
• Consider alternative loss functions and benchmark against these MasterThesis_BjoernMoelvig_qhm211_Final30sep_AsHandedIn.pdf