Super-sparse Learning in Similarity Spaces

Ambra Demontis, Marco Melis, Battista Biggio +2

In several applications, input samples are more naturally represented in terms of similarities between each other, rather than in terms of feature vectors. In these settings, machi…