Describe the bug
In the paper 'UniGNN: a Unified Framework for Graph and Hypergraph Neural Networks,' the input to the attention module of UniGAT is the vector obtained by concatenating the node vector and the hyperedge vector. However, during the implementation of the approach, in the forward function of UniGATConv, the input for atten_e is the hyperedge vector alone, rather than the vector formed by concatenating the node vector and the hyperedge vector.
I appreciate your contributions to such a valuable project, which has been immensely helpful to me in studying hypergraph techniques. Thank you for your attention.
Describe the bug
In the paper 'UniGNN: a Unified Framework for Graph and Hypergraph Neural Networks,' the input to the attention module of UniGAT is the vector obtained by concatenating the node vector and the hyperedge vector. However, during the implementation of the approach, in the forward function of UniGATConv, the input for atten_e is the hyperedge vector alone, rather than the vector formed by concatenating the node vector and the hyperedge vector.
I appreciate your contributions to such a valuable project, which has been immensely helpful to me in studying hypergraph techniques. Thank you for your attention.