Implementing DGCNN using DGL


if I would like to implement the DGCNN using DGL, do I have to use the DenseGraphConv layer instead of GraphConv before a SortPool layer?

The answer to this question depends on how dense your graphs will be. It seems to me that the graphs used in DGCNN are still normal graph datasets and probably not very dense. In this case, you will just need GraphConv.

1 Like