Multi-fidelity GPR
Subject of the issue
A multi-fidelity GPR model combines a small number of high-fidelity data points with a larger number of low-fidelity data points. The resulting model could potentially perform comparably to one trained exclusively on large amount of high-fidelity data.
Expected behaviour
A MultiFidelityGaussianProcessRegressor
class that has a similar interface to graphdot.model.gaussian_process.GaussianProcessRegressor
.
Edited by Yu-Hang "Maxin" Tang