There seems to be a problem with the code,1. In the creation of the adjacency matrix and 2. In the normalization of the adjacency matrix

[Zeng-CG]

1. Matrix of adjacency
   Not so:
   \begin{vmatrix}
   R & 0
   0 & 0
   \end{vmatrix}
   It goes like this:
   \begin{vmatrix}
   0 & R
   R^T & 0
   \end{vmatrix}

2. Normalization
   In the normalization operation, the occurrence of small values may lead to the problem of numerical overflow or insufficient precision
   After the normalization operation, this situation occurs:
   adj_t.to_dense()[:23566,:48123].sum(0)
   tensor([0.3358, 0.4679, 0.9746, ..., 0.0000, 0.0000, 0.0000])

Any questions are welcome to discuss