基于机器学习的智能路由解释方法

摘要：综述了基于机器学习的智能路由方法的进展，并提出了一种针对基于机器学习的智能路由技术的解释方法。该方法可以对神经网络等黑盒子技术的输出决策结果进行解释，支持几乎所有类型的智能路由算法。网络管理员可以利用该方法理解智能路由算法为何做出某些决策，并在此技术上进一步优化算法，排除故障，增强部署信心。

关键词：智能路由；图神经网络；超图；可解释性

Abstract: The recent advances in machine-learning-based intelligent routing algorithms are reviewed and a new interpretation method for machine-learning-based intelligent routing algorithms is proposed. This method can explain output decision results of black-box technologies such as neural networks. Network operators can therefore utilize such an interpretation method to understand the logic behind it. Further optimizations of the algorithm, debugging, and enhancing the confidence of deployment can be explored.

Keywords: intelligent routing; graph neural network; hypergraph; interpretability

在线PDF浏览： PDF