Hi. I'm having a trouble with understanding in this paper. On 6-2-1, Figure(a2, a4) shows that V = 90 shows the best F1 score. But the paper says that "The first is that there is most of the benefit is accomplished by starting v = 30 walks per node in both graph". Can't understand why..
Thank you so much for your great videoes. What is the difference between DeepWalk and GCN? I am confused about how to use them for real applications. Thanks a lot.
GCN learns the representation based on graph Laplacian, so the weights are kind of pre-defined by the graph structure; deepwalk learns the weights more "randomly" by the sampled walks.
keep on summarising papers that is awesome!
Thank you!
Nice work.
Hi. I'm having a trouble with understanding in this paper. On 6-2-1, Figure(a2, a4) shows that V = 90 shows the best F1 score. But the paper says that "The first is that there is most of the benefit is accomplished by starting v = 30 walks per node in both graph". Can't understand why..
Thank you so much for your great videoes. What is the difference between DeepWalk and GCN? I am confused about how to use them for real applications. Thanks a lot.
GCN learns the representation based on graph Laplacian, so the weights are kind of pre-defined by the graph structure; deepwalk learns the weights more "randomly" by the sampled walks.