This seems to be a way to identify and create a 'periodic table' of different machine learning approaches... importantly identifying 'gaps' between approaches which may yield new ways of doing various learning techniques.