Graph with homogeneous graph with multiple node features, multiple edge types, and edge weights to apply dual embedding

Questions
#1

Dear all,

I want to create a graph with homogeneous graph with multiple node features, multiple edge types, and edge weights to apply dual embedding (https://assets.amazon.science/d6/56/d03a00d14fd39c3486614e611e51/recommending-related-products-using-graph-neural-networks-in-directed-graphs.pdf).
e.g. directional graph, node = product, edge types = {co-view, co-purchase}, node feature = 256 dim, edge weight = (0, 1)

What is your suggestion for the graph creation? Could you please share similar example?

Thank you in advance for your help.

Best regards,
junhykim

#2

Here is an API for converting heterogeneous graphs into homogeneous graphs. After that, you can use RGCN for message passing.

1 Like
#3

Thank you for understanding modeling process!