
Geometric Graph Representations and Geometric Graph Convolutions for Deep Learning on ThreeDimensional (3D) Graphs
The geometry of threedimensional (3D) graphs, consisting of nodes and e...
read it

Gauge Equivariant Mesh CNNs: Anisotropic convolutions on geometric graphs
A common approach to define convolutions on meshes is to interpret them ...
read it

SchemaAware Deep Graph Convolutional Networks for Heterogeneous Graphs
Graph convolutional network (GCN) based approaches have achieved signifi...
read it

Molecular activity prediction using graph convolutional deep neural network considering distance on a molecular graph
Machine learning is often used in virtual screening to find compounds th...
read it

SWAG: Item Recommendations using Convolutions on Weighted Graphs
Recent advancements in deep neural networks for graphstructured data ha...
read it

Rubik: A Hierarchical Architecture for Efficient Graph Learning
Graph convolutional network (GCN) emerges as a promising direction to le...
read it

Hyperspectral Image Classification With ContextAware Dynamic Graph Convolutional Network
In hyperspectral image (HSI) classification, spatial context has demonst...
read it
DistanceGeometric Graph Convolutional Network (DGGCN)
The distancegeometric graph representation adopts a unified scheme (distance) for representing the geometry of threedimensional(3D) graphs. It is invariant to rotation and translation of the graph and it reflects pairwise node interactions and their generally local nature. To facilitate the incorporation of geometry in deep learning on 3D graphs, we propose a messagepassing graph convolutional network based on the distancegeometric graph representation: DGGCN (distancegeometric graph convolution network). It utilizes continuousfilter convolutional layers, with filtergenerating networks, that enable learning of filter weights from distances, thereby incorporating the geometry of 3D graphs in graph convolutions. Our results for the ESOL and FreeSolv datasets show major improvement over those of standard graph convolutions. They also show significant improvement over those of geometric graph convolutions employing edge weight / edge distance power laws. Our work demonstrates the utility and value of DGGCN for endtoend deep learning on 3D graphs.
READ FULL TEXT
Comments
There are no comments yet.