cancel
Showing results for 
Search instead for 
Did you mean: 

(node2vec,deepWalk..) VS(TransE,Rescal,ConvE..)

Hi everyone, I hope that you doing very well
I just start reading about Knowledge graph Embedding (KGE), I actually read a bunch of papers but I can't understand exactly whats the difference (if there is) between Node Representation and KGE, as well the difference between (node2vec,deepWalk..) AND (TransE,Rescal,ConvE..)
Thanks.

1 ACCEPTED SOLUTION

KG Embeddings and node embeddings are the same thing, conceptually - they represent each node in your graph as a fixed length vector.


KG embeddings like TransE, Rescal and ConvE are built explicitly for heterogeneous graphs - graphs with multiple node labels and multiple relationship types. Node2Vec, FastRP, and GraphSAGE weren't originally intended to be run on (or explicitly take into account) heterogeneous data. However, if you use GraphSAGE, our implementation does support multiple node labels (see our docs), or you can use FastRPExtended and encode the node labels as a property.

View solution in original post

2 REPLIES 2

KG Embeddings and node embeddings are the same thing, conceptually - they represent each node in your graph as a fixed length vector.


KG embeddings like TransE, Rescal and ConvE are built explicitly for heterogeneous graphs - graphs with multiple node labels and multiple relationship types. Node2Vec, FastRP, and GraphSAGE weren't originally intended to be run on (or explicitly take into account) heterogeneous data. However, if you use GraphSAGE, our implementation does support multiple node labels (see our docs), or you can use FastRPExtended and encode the node labels as a property.

Thank you so much ALICIA, by the way, I already watched your tutorial on youtube, he was a huge help for me, thanks again.

Nodes 2022
Nodes
NODES 2022, Neo4j Online Education Summit

All the sessions of the conference are now available online