Inverse operation in mean-centering and scaling in sampling

[yap231995]

I saw that the sampling in Lime lime_tabular.py is doing the following:

data = data * scale + instance_sample or data = data * scale + mean

May I know why would one want to do that with a scale and mean? Rather than just randomly sample from Gaussian distribution or just adding noise to the instance_sample when sample around instance_sample?

Thank you!