Besov Function Approximation and Binary Classification on
     Low-Dimensional Manifolds Using Convolutional Residual Networks

                 Hao Liu 1 Minshuo Chen 2 Tuo Zhao 2 Wenjing Liao 3

             Abstract                 The success of deep learning clearly demonstrates the great
                                  power of neural networks in representing complex data.
    Most of existing statistical theories on deep neural     In the past decades, the representation power of neural ...