Transfer Learning via `1 Regularization

        Masaaki Takada                  Hironori Fujisawa
       Toshiba Corporation           The Institute of Statistical Mathematics
       Tokyo 105-0023, Japan              Tokyo 190-8562, Japan
    masaaki1.takada@toshiba.co.jp             fujisawa@ism.ac.jp

                       Abstract
     Machine learning algorithms typically require abundant data under a stationary
     environment. However, environments are nonstationary in many real-world  ...