Algorithm 1. Extended bipartite graph convolutional network algorithm.

Input: Bipartite Graph G(V,E), Node Features X, Adjacency Matrix A, Network Depth L, Tolerance Error ε

Output: Node Embeddings Z

Preprocessing: Verify if A is the adjacency matrix of a bipartite graph, and normalize it;

Initialize node embeddings $H^{(0)}=X$;

for $l=1$ to L do

Compute normalized degree matrix $D^{-\frac{1}{2}}$;

Compute propagation matrix $D^{-\frac{1}{2}}A{D}^{-\frac{1}{2}}$;

Update the embedding matrix $H^{(l)}=\mathrm{R}\mathrm{e}\mathrm{L}\mathrm{U}(D^{-\frac{1}{2}}A{D}^{-\frac{1}{2}}{H}^{(l-1)}{W}^{(l)})$;

if $\parallel H^{(l)}-H^{(l-1)}\parallel <\epsilon$ then

break;

end

end

Output the final layer’s embedding matrix $Z=H^{(L)}$;

return Z;