1 Dicembre 2021, ore 11:00
Title: The Rare and the Unknown
Speaker: Dr.Stefano Cavuoti da Osservatorio Astronomico di Capodimonte
Abstract: The huge data volume produced by the new generation of astronomical facilities (both ground and space based) is clearly a goldmine, but they are posing new challenges at the same time. Identifying candidates of rare or peculiar objects is one of those. In the current work we decided to apply a completely unsupervised method (based on deep learning) directly to multi band images but without any aim to cluster objects together. The idea behind our method is quite intuitive: we transform the image of a source into a set of features that characterize the image itself and then we seek for other sources which are similar to the reference one. We present the method through an application to the identification of AGN candidates in the SDSS. The power of this method comes from the fact that, even if we have one single object of a kind, we can look for the most similar object in a dataset without the need of a large and well sampled training set and hence we can identify a list of candidates. For such reasons we think that it could be really helpful to deal with the datasets of large surveys to identify a list of candidates for rare objects.