Leveraging Non-uniformity in First-order Non-convex Optimization

         Jincheng Mei 1 2 * Yue Gao 1 * Bo Dai 2 Csaba Szepesvari 3 1 Dale Schuurmans 2 1

             Abstract                 in reinforcement learning (RL) (Agarwal et al., 2020), super-
                                  vised learning (SL) (Hazan et al., 2015), and deep learning
   Classical global convergence results for first-order      (Allen-Zhu et al., 2019). While gradient-based algorithms
   methods rely on uniform sm ...