this is my code when learning graph neural network hope this helpful to someone who learning ""if i describe something wrong am so sorry for that""
graph is the way to represent data that have a structure with a Vertex(Node) and edge. there are a way to represent a data like unweight(non-direction), weight and also bipatite graph. we can see the form of graph in a adjacency matix, edge_list, etc. in graph traditional method there are bag of degree, bag of graphlet, neighbor degree struture( degree is a node total number of connection), color refinement, etc all of these using static method to get graph information(node-level, edge-level, graph-level) .but the best way to learn represent of the graph is embedding and there are a method like node2vec and random walk but all of these is not looking into a node feature like random walk there random path and train to make it have a vector to close to each node if they have a short path so the new era of graph is a deeplearning method that really look into a node feature. this is really quick introduction from me if you interesting i reccommand to watch a lecture call cs224w by stanford in youtube this is really great free online source
this process is a message passing(aggreation) and pass through a liner layer if these process repeat N time it will mean having N layer of GCN all of these will output a embeding of Graph structure+feature and the last layer will be linear layer to classifire a node
- first embeding before train process
- last embeding after train process
