ljlilzu
commented
1 year ago More importantly, our method assumes that all layers have the same set of nodes. When performing link prediction in one layer, our method uses information not only in this layer but also in other layers. Our approach is not simply repeating the link prediction on each monolayer network. It takes into account the topological information of each layer, computes the similarity scores of each potential link on different layers, and then merges them using evidence theory.
Hello sir! I'm trying to use your model in multiplex network prediction and cite your model. While looking through your code, my colleagues and I are having conflicts on the model.
Lets say that our multiplex network consists of 3 layers. Looking at your paper and your code, I think that when your model makes link prediction on multiplex network, your model considers all three layers, therefore in case of layer 1, it does conduct link prediction on nodes that were not in layer 1.
However, my colleagues think that your model only considers each layer itself, therefore they are saying that is "same as running monoplex network three times, and printing out each result", there for it does not conduct link prediction on nodes that were not in layer 1.
Can you help me to explain my colleagues on this problem?