I'm wondering about the difference or the application of the different types of rescaling data.
So far, I'm aware that standardization assumes the data has a gaussian distribution. So if this is the case we should standardize and get values in normal distribution N~(0,1).
If our model does not has assumptions about the data distribution (like FCN or RandomForestClassifiers,...) and we don't know about the data distribution or it is just not gaussian distributed we should normalize the data. But here is my point, there are several methods to normalize e.g. using L2/L1-Norm of the vector (this is how tensorflow has implemented their standard normalization method) or using MinMaxScaling.
So when is normalization using either L1 or L2 norm recommended and when is MinMaxScaling the right choice?