gcn
gcn copied to clipboard
Effect of the number of epochs
Graph Convolution network models each node as an aggregation of the target node and its direct neighbors. By stacking more GCN layers we can access to further neighbors. I want to know, does training the model with some epochs on a one layer GCN has the effect similar to adding the second layer? or, Does second epoch of training means that the second neighbors' representations are affecting the target node representation?