GNNs are ideally suited for the Graphcore IPU. Learn how GNNs can benefit your business with blogs and case studies and find out how to get started today.Jump to Resources
Graph neural networks (GNNs) are AI models designed to derive insights from unstructured data described by graphs and an ideal fit for the Graphcore IPU, designed from the ground up for AI expressed as graphs. Unlike conventional CNNs, GNNs address the challenge of working with data in irregular domains. There are many applications for GNNs including molecular analysis, drug discovery, stock market prediction, social network analysis, traffic forecasting, recommender systems in e-commerce and much more.
Graph intelligence is a rapid growth segment within AI forecast to make up a sizeable percentage of the overall AI market in the coming years. Many businesses and organisations are discovering the potential of GNNs, and are applying them in a number of areas to drive innovation within their industry.
TGN: Temporal Graph Networks is a dynamic GNN model for training on the IPU using TensorFlow 1.
GIN: Graph Isomorphism Network (GIN) - a simple graph neural network to show that its discriminative/representational power is equal to the power of the Weisfeiler-Lehman (WL) graph isomorphism test.
An efficient algorithm for training deep and large Graph Convolutional Networks using TensorFlow 2.