About Gumbel_Generator in code and paper

bnusss / GGN

Gumbel Graph Network (GGN) : A General Deep Learning Framework for Network Reconstruction

69 stars 22 forks source link

About Gumbel_Generator in code and paper #2

Open xyk0058 opened 4 years ago

xyk0058 commented 4 years ago

in your code, aij and (1-aij) use gen_matrix = torch.randn (num_nodes, num_nodes, 2), is mean gen_matrix[:,:,0]+gen_matrix[:,:,1]=1 ? I'm confused about it. Please help me. Thanks!

3riccc commented 4 years ago

We don't have to set that limitation. gen_matrix[:,:,0] and gen_matrix[:,:,1] are just regular parameters and they are not mean for the probability of the edge. Rather, We use exp(g[:,:,0]) / (exp(g[:,:,0])+exp(g[:,:,1])) to represent the probability and they sum to 1 naturally.