Graph Neural Networks

I am currently learning some basic concepts related to Graph Neural Networks (GNN) and representation learning.

Node2Vec: simulations of biased random walks

I use this open-source node2vec implementation to simulate biased random walks in the node2vec algorithm. Following two videos illustrate the in-out parameter $q$’s role played in random walks with fixed return parameter $p$, i.e. a “ratio” of BFS vs. DFS.

• High in-out parameter $q$, Micro-view of neighbourhood, BFS

• Low in-out parameter $q$, Macro-view of neighbourhood, DFS