This can be done e.g. by looping over the parameters in the surrogate model, varying the current parameter randomly for each point in the test set and calculating 1-R2. This gives an indication of the sensitivity of the model to each parameter and for each output. @martinezcolanai made an implementation of this that we can reuse for harlow.
This can be done e.g. by looping over the parameters in the surrogate model, varying the current parameter randomly for each point in the test set and calculating
1-R2
. This gives an indication of the sensitivity of the model to each parameter and for each output. @martinezcolanai made an implementation of this that we can reuse for harlow.