研究人员开发了HiSE，一种专为异构图神经网络（HGNNs）设计的新型可解释模型。这种轻量级方法通过反映模型的语义层次结构，解决了在关键应用中解释HGNN决策的挑战。HiSE使用LASSO进行语义视图内的稀疏特征表示，并使用KL散度来统一这些视图的解释，在保真度和效率方面优于现有方法。 AI
影响 提高了复杂图神经网络的可解释性，这对于高风险应用至关重要。
排序理由 该集群包含一篇详细介绍解释机器学习模型新方法的论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2606.03495v1 Announce Type: new Abstract: Heterogeneous graph neural networks (HGNNs) have demonstrated remarkable performance in modeling complex relational data, however their interpretability in high-stakes applications remains a critical challenge. Existing explanation …
Heterogeneous graph neural networks (HGNNs) have demonstrated remarkable performance in modeling complex relational data, however their interpretability in high-stakes applications remains a critical challenge. Existing explanation methods suffer from two major limitations: on th…