A question about the code

[fireppGo]

Sorry, there are some doubts in reading the paper and code. In the dcrnn_cell.py, start on line 161, why is this a concat operation. Is this step reflected in the paper? And how to understand x2 = 2 * tf.sparse_tensor_dense_matmul(support, x1) - x0. To best of my knowledge, diffusion convolution is a summation operation in the paper. And the paper doesn't seem to mention the operation of stacking. Looking forward to your answer, thanks.