The full exploitation of the next generation of large scale photometric surveys depends heavily on our ability to provide reliable classification and redshifts  based solely on photometric data. In preparation for this scenario, there has been many attempts to approach  the supernova photometric classification and the photometric redshift estimation issues. Although different methods present different degree of success, text-book machine learning methods fail to address the crucial issue of lack of representativeness between spectroscopic (training) and photometric (target) samples. In this talk I will show how Adaptive Learning can be used to construct a recommendation systems specially design for the astronomical case and ensure optimal exploration of spectroscopic resources. I will also illustrate how this crucial issue lies in the boundary between machine learning and statistics, helping us to appreciate the connection between these two fields.