You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
we've been running node2vec and noticed that if a graph does not have a node 0, random walks that have no other neighbour give a vector that is filled up with 0s for the remaining steps of the walk.
We saw that this is because the random walk breaks with no other neighbours and 0 is considered the root node.
In a non-singularly rooted graph, if we wanted to consider as roots all seed nodes (each one in its own random walk), wouldn't that be a more faithful representation of the graph ? Since if there is no 0 node, a node is added, and if there is a 0 node in the network then walks are randomly returning to it, biasing the embeddings from word2vec ?
Thanks in advance if you have an answer and/or have already tried it!
The text was updated successfully, but these errors were encountered:
Hi,
we've been running node2vec and noticed that if a graph does not have a node 0, random walks that have no other neighbour give a vector that is filled up with 0s for the remaining steps of the walk.
We saw that this is because the random walk breaks with no other neighbours and 0 is considered the root node.
In a non-singularly rooted graph, if we wanted to consider as roots all seed nodes (each one in its own random walk), wouldn't that be a more faithful representation of the graph ? Since if there is no 0 node, a node is added, and if there is a 0 node in the network then walks are randomly returning to it, biasing the embeddings from word2vec ?
Thanks in advance if you have an answer and/or have already tried it!
The text was updated successfully, but these errors were encountered: