Provable Meta-Learning of Linear Representations

                  Nilesh Tripuraneni 1 Chi Jin 2 Michael I. Jordan 1

              Abstract               ranging from deep reinforcement learning (Baevski et al.,
   Meta-learning, or learning-to-learn, seeks to de-      2019) to natural language processing (Ando & Zhang, 2005;
   sign algorithms that can utilize previous experi-      Liu et al., 2019). Accordingly, researchers have begun to
   ence to rapidly learn new skills or adapt to new   ...