GeniePath: Graph Neural Networks with Adaptive Receptive Paths
This work addresses the challenge of scalable and adaptive graph learning for researchers and practitioners in machine learning, representing an incremental improvement over existing methods.
The authors tackled the problem of learning adaptive receptive fields in graph neural networks for permutation invariant graph data, achieving state-of-the-art results on large graphs in both transductive and inductive settings.
We present, GeniePath, a scalable approach for learning adaptive receptive fields of neural networks defined on permutation invariant graph data. In GeniePath, we propose an adaptive path layer consists of two complementary functions designed for breadth and depth exploration respectively, where the former learns the importance of different sized neighborhoods, while the latter extracts and filters signals aggregated from neighbors of different hops away. Our method works in both transductive and inductive settings, and extensive experiments compared with competitive methods show that our approaches yield state-of-the-art results on large graphs.