A Polynomial Architecture-Attribution Co-Design Framework for Exact Aumann-Shapley Attribution in GNNs
一种用于图神经网络中精确奥曼-夏普利归因的多项式架构-归因协同设计框架
机构 * Institute of Artificial Intelligence Innovation and Industry, Fudan University(复旦大学人工智能创新与产业研究院) ; Shanghai Academy of AI for Science(上海人工智能科学研究院) ; Human Phenome Institute, Fudan University(复旦大学人类表型组研究院) ; School of Information and Communication Engineering, Communication University of China(中国传媒大学信息与通信工程学院)
AI总结 研究图神经网络的特征级和节点级解释,提出APEX框架,通过PolyGIN使归因积分可精确计算,实验表明该框架能保持预测性能,提高归因保真度并减少评估次数。
Comments It has 23 pages