For Course 1 of Machine Learning Specialization:

Why are we not verifying assumptions of (a) linear regression and (b) logistic regression, in particular: (a) normally distributed residuals (losses), constant variance of residuals, etc and (b) multicolinearity, independent observations, etc.

Is this not done in Data Science? Or is it in a future module?

Thank you in advance.

This question comes up fairly often from people who have strong backgrounds in statistics.

Those are more in line with “human learning” processes.

Since Machine Learning tries to make the learning as automated as possible, those methods aren’t commonly used.

It’s also possible that in some cases, those practices happen behind the scenes, in preparing a data set, before any machine learning happens.