Do you have any references for the approach that you recommended? You're saying that differentiation (not even just numerical differentiation) is "not something

Do you have any references for the approach that you recommended? You're saying that differentiation (not even just numerical differentiation) is "not something you want to apply to real-world data" ... This is surprising since differentiation obviously is used in the real-world in some very critical applications.