GP evaluation

[vatsgo]

Evaluation pf the GP quality.

Added the GP evaluation for the North sea wave crest heights use case.
The Nootebook describes the assumptions as well as how to interpret the results. A summary of the plots is here :

[am-kaiser]

In my review I have stayed on high level perspective so far and haven't yet looked at the code in detail. @swinter1 will take over from here.

[am-kaiser]

Please add a description of what you are doing and why at the top of the notebook. The idea is that someone else who is reading this in 1 year still understands what is going on.

[am-kaiser]

Specify what all variables mean, e.g. X, Y, theta.

[am-kaiser]

experiment.trials
└── Trial (index 0)
└── Arm "0_0"
├── arm.parameters = {"Hs": 4.2, "Tp": 8.7} → X
└── trial.run_metadata["simulation_result"]
→ LocalMetadataMetric reads it
→ df row: arm_name="0_0", metric_name="loc"
mean=12.3, sem=0.4 # mean → Y, sem² → Yvar

Is not displayed well. Maybe you can find a better way of describing this. As a one liner it is hard to understand.

[am-kaiser]

Please add more explanations what you are doing now and why. For example "extract_training_tensors" comes a bit out of thin air. It's great that you have added a docstring but in a notebook I would expect a short explanation before the code.

[am-kaiser]

Please split up your code segments more. E.g. the segment of run_loo is very long. If you split it up into smaller chuncks, e.g. one per function/ class, and add some high level explanation then it would be much easier to understand what is going on.