One reason I think people don’t do error analysis is there’s no recipe for it. You just have to be smart and resourceful and understand the domain. In contrast for all the previous steps of the ML pipeline there are standard recipes and many steps are even automatable.
This is good!
One reason I think people don’t do error analysis is there’s no recipe for it. You just have to be smart and resourceful and understand the domain. In contrast for all the previous steps of the ML pipeline there are standard recipes and many steps are even automatable.