Various approaches have been proposed for static graph embedding.
Examples include SVD based models, Random-walk based models, deep autoencoder model. Existing works which focus on dynamic embeddings apply static embedding algorithms to each snapshot of the dynamic graph and then rotationally align the resulting static embeddings across time steps.
Existing model have limitations
Stability: The embedding generated by static methods is not stable.
Growing Graphs: New nodes can be introduced into the graph and create new links to existing nodes as the dynamic graph grows in time.
Scalability: Learning embeddings independently for each snapshot leads to running time linear in the number of snapshots.
DynGEM is a dynamic graph neural network models for growing graph that has auto encoder structures.
Various approaches have been proposed for static graph embedding.
Examples include SVD based models, Random-walk based models, deep autoencoder model. Existing works which focus on dynamic embeddings apply static embedding algorithms to each snapshot of the dynamic graph and then rotationally align the resulting static embeddings across time steps.
Existing model have limitations
Stability: The embedding generated by static methods is not stable.
Growing Graphs: New nodes can be introduced into the graph and create new links to existing nodes as the dynamic graph grows in time.
Scalability: Learning embeddings independently for each snapshot leads to running time linear in the number of snapshots.
DynGEM is a dynamic graph neural network models for growing graph that has auto encoder structures.