IEEE P2941.4

New IEEE Standard - Active - Draft.
This standard defines the representation method, model, and Application Programming Interface (API) for graph neural network (GNN), including graph data representation, GNN model specification, and methods for compressing and accelerating GNN models. It aims to establish unified interfaces and guidelines to enhance interoperability, efficiency, and ease of deployment for GNN applications across diverse platforms and computational environments.

This standard specifies a representation and an application programming interface (API) for graph neural network models used for various computing needs.
It provides
the representation and basic operations of graph data,
the representation, interface definitions, and compression methods for graph neural network models, and the computing framework for graph neural networks.
This standard aims to provide researchers and developers with unified graph data, graph neural network representation, and interface definitions of various graph neural network models and compression technologies. Thereby it can promote the research, development, testing and evaluation of graph neural network models, as well as their applications in different platforms and domains.