利用随机噪声的等变性普遍提升图神经网络
Using Random Noise Equivariantly to Boost Graph Neural Networks Universally
AI总结：
本文提出等变噪声图神经网络ENGNN，通过理论分析随机噪声引入GNN的样本复杂度问题，并利用噪声对称性提升节点、链路、子图及图级任务的表达能力与泛化性能。
摘要：
图神经网络（GNNs）的最新进展探索了将随机噪声作为输入特征以增强其在多种任务上的表达能力的潜力。然而，简单地引入噪声可能会降低性能，而为特定任务量身定制以利用噪声的架构虽然表现出色，却缺乏广泛的适用性。本文通过建立一个理论框架来解决这些问题，该框架阐明了在未经精心设计的情况下将随机噪声引入GNNs时所增加的样本复杂度。我们进一步提出了等变噪声图神经网络（ENGNN），这是一种新颖的架构，利用噪声的对称性质来降低样本复杂度并增强泛化能力。我们的实验表明，以等变方式使用噪声能显著提升节点级、链路级、子图级和图级任务的性能，并达到与为特定任务设计的模型相当的性能，从而提供了一种在各种图任务中提升表达能力的通用方法。
英文摘要：
Recent advances in Graph Neural Networks (GNNs) have explored the potential of random noise as an input feature to enhance expressivity across diverse tasks. However, naively incorporating noise can degrade performance, while architectures tailored to exploit noise for specific tasks excel yet lack broad applicability. This paper tackles these issues by laying down a theoretical framework that elucidates the increased sample complexity when introducing random noise into GNNs without careful design. We further propose Equivariant Noise GNN (ENGNN), a novel architecture that harnesses the symmetrical properties of noise to mitigate sample complexity and bolster generalization. Our experiments demonstrate that using noise equivariantly significantly enhances performance on node-level, link-level, subgraph, and graph-level tasks and achieves comparable performance to models designed for specific tasks, thereby offering a general method to boost expressivity across various graph tasks.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
