- 10,000 randon generate graph
- trainset
- networkx.generators.random_graphs.powerlaw_cluster_graph(n=200, m=4, p=0.05)
- 100 randon generate graph
- validset
- networkx.generators.random_graphs.powerlaw_cluster_graph(n=200, m=4, p=0.05)
- loss: BCELoss
- GT: sigmoid(cc[det] - cc[src])
- batch_size: 32
- epochs: 100
- Save model based on best validation evaluation of top 10