GAE and Node2Vec for edge analysis
Today's tutorial shows how to use previous models for edge analysis.
We first use Graph Autoencoder to predict the existence of an edge between nodes, showing how simply changing the loss function of GAE, can be used for link prediction.
Later, we propose the use of Node2Vec for edge-label prediction. In particular, we build a node embedding, then we compute the edge embedding as the mean of the nodes embedding of the link.
Then, we use the node embedding and Random Forest Classifier for edge label prediction.
Download the material of the lecture here.