Paper Notes: Heterogeneous Graph Transformer (HGT)
--- Traditional Graph Neural Networks (GNNs) are primarily designed for homogeneous graphs, where all nodes and edges share the same type. Consequently, they cannot effectively represent heterogene…
Graph Learning
This section collects graph-based neural models and related methods.
Articles
Relational Inductive Biases, Deep Learning, and Graph Networks
--- - To reach human-like intelligence, combinatorial generalization is the primary goal. - Definition: use a finite set of known building blocks to construct an unbounded set of inferences. - This…