FIT-GNN: Faster Inference Time for GNNs that 'FIT' in Memory Using Coarsening

arXiv stat.ML / 4/13/2026

💬 OpinionIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper argues that while graph coarsening and related techniques help GNN training, prior work has not sufficiently reduced the computational and memory costs during GNN inference.
It introduces FIT-GNN, which applies graph coarsening at inference time to reduce computational burden, and proposes two variants: Extra Nodes and Cluster Nodes.
FIT-GNN targets graph-level tasks, demonstrating results for graph classification and graph regression rather than only node-level workloads.
Experiments on multiple benchmark datasets show orders-of-magnitude faster single-node inference time versus traditional methods.
The approach also significantly lowers memory usage, making efficient training and inference feasible on low-resource devices without major performance loss compared to baselines.

Abstract

Scalability of Graph Neural Networks (GNNs) remains a significant challenge. To tackle this, methods like coarsening, condensation, and computation trees are used to train on a smaller graph, resulting in faster computation. Nonetheless, prior research has not adequately addressed the computational costs during the inference phase. This paper presents a novel approach to improve the scalability of GNNs by reducing computational burden during the inference phase using graph coarsening. We demonstrate two different methods -- Extra Nodes and Cluster Nodes. Our study extends the application of graph coarsening for graph-level tasks, including graph classification and graph regression. We conduct extensive experiments on multiple benchmark datasets to evaluate the performance of our approach. Our results show that the proposed method achieves orders of magnitude improvements in single-node inference time compared to traditional approaches. Furthermore, it significantly reduces memory consumption for node and graph classification and regression tasks, enabling efficient training and inference on low-resource devices where conventional methods are impractical. Notably, these computational advantages are achieved while maintaining competitive performance relative to baseline models.