用循环网络替代图神经网络聚合，提升算法推理能力
Recurrent Aggregators in Neural Algorithmic Reasoning
- 用循环神经网络替代传统图神经网络的聚合方式
- 在堆排序和快速选择任务上达到87%的F1分数，领先现有方法
- 适合需要序列依赖建模的算法推理任务
神经算法推理（NAR）旨在设计模仿经典算法计算的神经网络。当前图神经网络（GNN）因消息传递框架和置换等变性被广泛采用。本文挑战这一设计，将等变聚合函数替换为循环神经网络。尽管看似反直觉，但在节点具有自然顺序的任务中具备合理基础——这在CLRS-30等经典推理基准中常见。所提出的循环神经算法推理（RNAR）模型在该类任务上表现优异，同时对其他任务也具有良好泛化能力。其显著成果在于在堆排序与快速选择任务上取得领先，尤其在快速选择任务中实现均值微平均F1分数87%，这是当前神经算法推理器的重大挑战。
原文摘要 · Abstract (English)
Neural algorithmic reasoning (NAR) is an emerging field that seeks to design neural networks that mimic classical algorithmic computations. Today, graph neural networks (GNNs) are widely used in neural algorithmic reasoners due to their message passing framework and permutation equivariance. In this extended abstract, we challenge this design choice, and replace the equivariant aggregation function with a recurrent neural network. While seemingly counter-intuitive, this approach has appropriate grounding when nodes have a natural ordering -- and this is the case frequently in established reasoning benchmarks like CLRS-30. Indeed, our recurrent NAR (RNAR) model performs very strongly on such tasks, while handling many others gracefully. A notable achievement of RNAR is its decisive state-of-the-art result on the Heapsort and Quickselect tasks, both deemed as a significant challenge for contemporary neural algorithmic reasoners -- especially the latter, where RNAR achieves a mean micro-F1 score of 87%.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。