If I understood, the first 3 points can be considered as outliers. Typically, you do want to ignore them, do to them being mistakes in measurements or special c

If I understood, the first 3 points can be considered as outliers. Typically, you do want to ignore them, do to them being mistakes in measurements or special cases. In any case, handling of the special cases does not do well for generalization of the algorithms. So I would be against those points. 4th points is true, except that it's more of a zen advice then actual useful information. 5th is a comercial. Not a very good read, and I generally found that no machine learning knowledge gets explained in 5 short paragraphs with no technical terms (except "sigma", which I guess is standard deviation).