Local-to-global Perspectives on Graph Neural Networks
It addresses foundational challenges in graph-structured data processing for machine learning, but appears incremental in its contributions.
This thesis tackles the problem of understanding and improving graph neural networks (GNNs) by categorizing them into local and global types, studying convergence properties, connecting these types, and applying local methods to graph coarsening.
This thesis presents a local-to-global perspective on graph neural networks (GNN), the leading architecture to process graph-structured data. After categorizing GNN into local Message Passing Neural Networks (MPNN) and global Graph transformers, we present three pieces of work: 1) study the convergence property of a type of global GNN, Invariant Graph Networks, 2) connect the local MPNN and global Graph Transformer, and 3) use local MPNN for graph coarsening, a standard subroutine used in global modeling.