IGLU: Efficient GCN Training via Lazy Updates

TL;DR

IGLU is a novel method for training multi-layer GCNs that caches intermediate computations to enable lazy updates, significantly reducing compute costs while maintaining or improving accuracy.

Contribution

IGLU introduces a caching-based lazy update mechanism for GCN training that reduces computation and maintains convergence despite bounded bias.

Findings

Up to 88% reduction in compute cost

Achieves up to 1.2% better accuracy

Converges to a first-order saddle point

Abstract

Training multi-layer Graph Convolution Networks (GCN) using standard SGD techniques scales poorly as each descent step ends up updating node embeddings for a large portion of the graph. Recent attempts to remedy this sub-sample the graph that reduces compute but introduce additional variance and may offer suboptimal performance. This paper develops the IGLU method that caches intermediate computations at various GCN layers thus enabling lazy updates that significantly reduce the compute cost of descent. IGLU introduces bounded bias into the gradients but nevertheless converges to a first-order saddle point under standard assumptions such as objective smoothness. Benchmark experiments show that IGLU offers up to 1.2% better accuracy despite requiring up to 88% less compute.