Current browse context:
cs.LG
Change to browse by:
References & Citations
Computer Science > Machine Learning
Title: Neural Message Passing on High Order Paths
(Submitted on 24 Feb 2020)
Abstract: Graph neural network have achieved impressive results in predicting molecular properties, but they do not directly account for local and hidden structures in the graph such as functional groups and molecular geometry. At each propagation step, GNNs aggregate only over first order neighbours, ignoring important information contained in subsequent neighbours as well as the relationships between those higher order connections. In this work, we generalize graph neural nets to pass messages and aggregate across higher order paths. This allows for information to propagate over various levels and substructures of the graph. We demonstrate our model on a few tasks in molecular property prediction.
Submission history
From: Daniel Flam-Shepherd [view email][v1] Mon, 24 Feb 2020 17:58:02 GMT (126kb,D)
Link back to: arXiv, form interface, contact.