KD3A: Unsupervised Multi-Source Decentralized Domain Adaptation via
             Knowledge Distillation

         Hao-zhe Feng 1 Zhaoyang You 2 Minghao Chen 1 Tianye Zhang 1 Minfeng Zhu 1
                    Fei Wu 2 Chao Wu 3 Wei Chen 1

             Abstract                 to train the model. However, due to the domain shift,
                                  naively combining different datasets often results in unsat-
   Conventional unsupervised multi-source domain         isfying performan ...