Autoregressive Models for Sequences of Graphs
This work addresses a domain-specific problem for researchers in graph-based machine learning, representing an incremental advancement by generalizing autoregressive models to graph sequences.
The paper tackles the problem of modeling sequences of graphs with variable topologies and attributes by proposing an autoregressive model that uses a graph neural network to predict the next graph, achieving significantly better performance than four baselines on synthetic graph-generating processes.
This paper proposes an autoregressive (AR) model for sequences of graphs, which generalises traditional AR models. A first novelty consists in formalising the AR model for a very general family of graphs, characterised by a variable topology, and attributes associated with nodes and edges. A graph neural network (GNN) is also proposed to learn the AR function associated with the graph-generating process (GGP), and subsequently predict the next graph in a sequence. The proposed method is compared with four baselines on synthetic GGPs, denoting a significantly better performance on all considered problems.