Results for K-folds cross validation and leave-one-out RMSE seem incoherent

Summary

Results for validation of surrogate model seem incoherent.

Gemseo version

develop

Platform info

Windows 10

Environment info

standard python env produced from tox

Steps to reproduce

Run the script: surrogate.py

What is the current bug behavior?

The surrogate model produced on the learning dataset fits very well the testing dataset.

This is the last output of the script:

Learning RMSE = 0.003228079021716151
Test RMSE = 0.029124582305771165
Generalization RMSE = 0.15076326538286983
K-folds cross-validation RMSE = 22.814476006715417
Leave-one-out RMSE = 28.251585950274567

The first three error measurements are coherent with the fact that the surrogate is accurate. However, the last two show large errors. It may come from the fact that I do not fully understand what these criteria are, but the error value seem very large compared to the accuracy of the surrogate.

What is the expected correct behavior?

All five criteria shown above have low values.