Decide on high-level interface

[mastoffel]

We want a generic interface to

• take input and output of a simulation
• fit and evaluate several emulators
• return best emulator model

Therefore, it'd be good to decide on a set of high-level functions that minimise thinking time for users, a la scikit-learn.

[mastoffel]

First thoughts:

experimental_design.py -> samplers (e.g. latin hypercube) emulators.py -> models (e.g. Gaussian Process, Random Forests) compare.py -> high level interface to compare models and provide metrics

The simplest interface would be to have a single function:

compare(X, y)

• X are simulation inputs and y are simulation outputs
• runs k-fold cross validation on all models
• model parameters could be either automatic (chosen based on data structure) or user-specified