Accumulated Decoupled Learning with Gradient Staleness Mitigation for
           Convolutional Neural Networks

      Huiping Zhuang 1 Zhenyu Weng 1 Fulin Luo 1 Kar-Ann Toh 2 Haizhou Li 3 Zhiping Lin 1

             Abstract                 efficiency, the decoupled learning (Jaderberg et al., 2016)
                                  emerges by addressing one or more of these lockings.
   Gradient staleness is a major side effect in decou-
   pled learning when training convolutional neural       ...