# Training a model to recognize difference between simulations

I have been facing a problem where I have two simulations. One quite heavy, but potent and able to generalize well the real scenario, and the other which has a low fidelity. Both simulation frameworks have the same inputs and the same outputs.

Therefore, I was wondering if I could somehow use machine learning to create a connection between the two models. Like, I would create let's say, a dataset of 40 data points using the same inputs with the low fidelity and the high fidelity, then I would train a model, and with that trained model, I could use it to improve the results of the low fidelity framework to use it instead of using the computational costly high fidelity framework. However, I was wondering how could data be done.

So, supposing I have XL1, XL2, XL3 as inputs and YL1 as output of my low fidelity model, and then XH1,XH2,XH3 as inputs and YH1, is it possible to in such a way try new values of the low fidelity XL'1, XL'2,XL'3 to predict what would be the correspondent YH'1?