Learning from Biased Data: A Semi-Parametric Approach

         Patrice Bertail * 1 Stephan Clemencon * 2 Yannick Guyonvarch * 2 Nathan Noiry * 2

             Abstract                 distribution P and ` : Z × Θ → R+ is a given loss func-
                                  tion. In order to solve approximately the risk minimization
   We consider risk minimization problems where
                                  problem
   the (source) distribution PS of the training obser-
                   ...