Summary
Many graph-based AI models trained on small graphs struggle when used on bigger ones because they can only look a few steps away. The authors study a way to reuse a learned local rule multiple times to gather information beyond what was seen during training. They show that by repeating this small learned step, the model can effectively understand larger graphs without needing to retrain the whole system. To improve this reuse, they propose a new method that keeps the main model fixed but learns a small controller to guide it, improving accuracy with fewer parameters. Their tests show this approach works well on artificial problems and chip design tasks.
What this means in practice
- •For chip design engineers: Improve placement algorithms by enabling pretrained graph models to handle larger chip layouts without complete retraining.
- •For network infrastructure planners: Use prelearned local rules repeatedly to analyze and plan larger communication graphs efficiently as networks grow.
Abstract
Generalizing from small graphs to larger ones can require computation beyond the depth of a neural model used during training. We study target tasks that admit a common local iterative computation and whose effective dependency radius grows with graph size. The central idea is to learn a local update rule through GNN pretraining, then repeatedly apply the same learned processor beyond the training depth to extend information reach on larger graphs without retraining. Calling such parameter-shared architecture a \emph{looped GNN}, we categorize six schemes for looped GNNs to separate the effects of depth extension, fine-tuning, and freeze-and-steer adaptation. Our analysis establishes fixed-depth information limits and sufficient conditions under which accurate iterative reuse overcomes them. To improve beyond unadapted reuse, we introduce \emph{Loop-Freeze-Steer} (LFS), which freezes the pretrained processor and learns a small controller to steer its repeated application, using much fewer trainable parameters than full fine-tuning. Our finite-sample analysis gives sufficient conditions for adaptation to improve performance on larger graphs. Experiments on synthetic tasks and chip placement evaluate depth extension and adaptation, highlighting LFS's accuracy--trainable-parameter trade-off for size generalization.