Scalability of Partition and Code for large networks

Ascertain whether the Partition and Code graph compression method can scale to network graphs with millions of nodes and edges, establishing its computational feasibility and performance at large scale.

Background

Partition and Code is a machine learning-based approach that decomposes graphs into subgraphs and learns a dictionary code. Although it reports good compression on small graph datasets, the thesis questions its applicability at the scale of modern networks with millions of nodes and edges.

Determining the scalability of Partition and Code would clarify its role among large-graph compression methods and guide future algorithmic and model-design choices for network data.