Non-Negative Bregman Divergence Minimization
               for Deep Direct Density Ratio Estimation

                       Masahiro Kato 1 Takeshi Teshima 2

              Abstract               the density ratio when anomaly-free unlabeled test data are
                                available (Hido et al., 2008).
    Density ratio estimation (DRE) is at the core of
    various machine learning tasks such as anomaly     Among the various approaches to DRE, we focus on the
    detection and d ...