HyperGraphX: Graph Transductive Learning with Hyperdimensional Computing and Message Passing

TL;DR

HyperGraphX ( exttt{HDGC}) introduces a novel hyperdimensional computing-based graph transductive learning algorithm that outperforms existing GNNs and hyperdimensional methods in accuracy and speed, with promising energy efficiency on specialized hardware.

Contribution

The paper presents exttt{HDGC}, a new algorithm combining graph convolution with hyperdimensional computing, achieving superior accuracy and speed over existing methods.

Findings

exttt{HDGC} outperforms major GNNs and hyperdimensional methods in accuracy.

exttt{HDGC} is significantly faster than exttt{GCNII} and HDGL on the same GPU platform.

exttt{HDGC} is expected to have high energy efficiency on neuromorphic hardware.

Abstract

We present a novel algorithm, \hdgc, that marries graph convolution with binding and bundling operations in hyperdimensional computing for transductive graph learning. For prediction accuracy \hdgc outperforms major and popular graph neural network implementations as well as state-of-the-art hyperdimensional computing implementations for a collection of homophilic graphs and heterophilic graphs. Compared with the most accurate learning methodologies we have tested, on the same target GPU platform, \hdgc is on average 9561.0 and 144.5 times faster than \gcnii, a graph neural network implementation and HDGL, a hyperdimensional computing implementation, respectively. As the majority of the learning operates on binary vectors, we expect outstanding energy performance of \hdgc on neuromorphic and emerging process-in-memory devices.