Page 208 - 《软件学报》2024年第4期

1786 软件学报 2024 年第 35 卷第 4 期

(a) p c=1 (b) p c=0.8
图 4 消融实验结果

4 总 结

本文提出了基于强化联邦图神经网络的个性化公共安全突发事件检测方法 PPSED, 此方法解决了在传统
全局联邦方法中由于非独立同分布(Non-IID)数据分布和隐私约束所带来的挑战, 通过使客户端能够协同训练
个性化且鲁棒的模型, 来处理本地的突发事件检测任务. PPSED 方法包含几种创新的设计策略. 设计了联邦
公共安全突发事件检测模型结构与梯度量化方法, 利用基于图采样的 minibatch 机制结合 GraphSage 来构建本
地模型, 并且选择客户端梯度进行量化, 减轻 Non-IID 数据的影响, 同时减少梯度通信消耗. 设计了一个基于
随机图嵌入的客户端状态感知方法, 以在保证隐私的同时, 更好地保留客户端模型的有价值梯度信息. 设计
了强化联邦图神经网络的个性化梯度聚合和量化策略, 以加强联邦图神经网络. 通过采用 DDPG, 拟合个性化
的联邦学习梯度聚合加权策略, 并根据权重决定是否对梯度进行量化, 从而在模型性能和通信压力之间找到
平衡. 在微博平台收集的公共安全数据集和 3 个其他公开的图数据集上进行了广泛的实验, 结果表明了所提
方法的有效性.

References:
[1] Cao Y, Peng H, Wu J, et al. Knowledge-preserving incremental social event detection via heterogeneous GNNs. In: Proc. of the
Web Conf. 2021. 3383−3395.
[2] Liu Z, Yang Y, Huang Z, et al. Event early embedding: Predicting event volume dynamics at early stage. In: Proc. of the 40th Int’l
ACM SIGIR Conf. on Research and Development in Information Retrieval. 2017. 997−1000.
[3] Fedoryszak M, Frederick B, Rajaram V, et al. Real-time event detection on social data streams. In: Proc. of the 25th ACM
SIGKDD Int’l Conf. on Knowledge Discovery & Data Mining. 2019. 2774−2782.
[4] Aggarwal CC, Subbian K. Event detection in social streams. In: Proc. of the 2012 SIAM Int’l Conf. on Data Mining. Society for
Industrial and Applied Mathematics, 2012. 624−635.
[5] Peng H, Li J, Song Y, et al. Streaming social event detection and evolution discovery in heterogeneous information networks. ACM
Trans. on Knowledge Discovery from Data, 2021, 15(5): 1−33.
[6] Cui W, Du J, Wang D, et al. Extended search method based on a semantic hashtag graph combining social and conceptual
information. World Wide Web, 2019, 22: 2589−2610.
[7] Hamilton W, Ying Z, Leskovec J. Inductive representation learning on large graphs. In: Advances in Neural Information Proccesing
Systems, Vol.30. 2017
[8] Kipf TN, Welling M. Semi-supervised classification with graph convolutional networks. arXiv:1609.02907, 2016.
[9] Bo D, Wang X, Liu Y, et al. A survey on spectral graph neural networks. arXiv:2302.05631, 2023.
[10] Baek J, Kang M, Hwang SJ. Accurate learning of graph representations with graph multiset pooling. arXiv:2102.11533, 2021.
[11] Wu Z, Pan S, Chen F, et al. A comprehensive survey on graph neural networks. IEEE Trans. on Neural Networks and Learning
Systems, 2021, 32(1): 4−24.
[12] Perozzi B, Al-Rfou R, Skiena S. DeepWalk: Online learning of social representations. In: Proc. of the 20th ACM SIGKDD Int’l
Conf. on Knowledge Discovery and Data Mining. 2014. 701−710.