The mathematical framework of approximate entropic learning introduced very recently promises to provide robust, cheap and efficient ways of machine learning in the so called "small data" regime, when the underlying learning ...