基于隐私保护机制的辐射光源衍射图像筛选  

Diffraction Image Screening of Radiation Facilities Based on Privacy Protection Mechanism

作　　者：许康 祝永新 吴波[2] 郑小盈[2] 陈凌曜 Xu Kang;Zhu Yongxin;Wu Bo;Zheng Xiaoying;Chen Lingyao(School of Integrated Circuits,University of Chinese Academy of Sciences,Beijing 100049,China;Shanghai Advanced Research Institute,Chinese Academy of Sciences,Shanghai 201210,China;Shanghai Information Technology Research Center,Shanghai 201210,China)

机构地区：[1]中国科学院大学集成电路学院,北京100049 [2]中国科学院上海高等研究院,上海201210 [3]上海信息技术研究中心,上海201210

出　　处：《激光与光电子学进展》2023年第10期206-214,共9页Laser & Optoelectronics Progress

基　　金：国家自然科学基金(U2032125);科技部SKA专项(2020SKA0120202)。

摘　　要：同步辐射光源产生超高速的衍射图像数据流,需要通过数据筛选降低数据传输和存储的压力。但互相竞争的研究小组不愿意分享数据,现有基于深度学习的筛选方法难以应对隐私保护下有效训练的挑战,因此首次将联邦学习技术应用在辐射光源衍射图像筛选中,通过数据和模型分离,实现隐私保护下的训练数据增广。提出筛选方法Federated Kullback-Leibler(FedKL),基于改进的KL散度和数据量权重,对全局模型更新进行改进,在获得高准确率的同时降低算法的复杂度,满足高速数据流高精度处理要求。针对异地光源多中心数据同步训练的困难,又提出同步和异步相结合的混合训练方式,在不降低模型识别准确率的同时,显著提升了模型的训练速度。在光源CXIDB-76公开数据集上的实验结果表明,相比FedAvg,FedKL能够提升准确率和F1分数,分别提升了25.2个百分点和0.419。Synchrotron radiation facilities generate ultrahighspeed diffraction image data streams,which require data screening to reduce the pressure on data transmission and storage.However,competing research groups are reluctant to share such data,and existing deep learningbased screening methods cannot easily achieve effective training under privacy protection.Therefore,for the first time,this study applies the federated learning technology to the screening of radiation source diffraction images,and training data augmentation under privacy protection is realized by separating the data and the model.The Federated KullbackLeibler(FedKL)screening method is also proposed to improve the global model update based on KullbackLeibler divergence and data volume weights,thus reducing the complexity of the algorithm while obtaining high accuracy;further,this satisfies the highprecision processing requirements for highspeed data streams.To address the difficulties encountered in data synchronization training for multiple centers of remote light sources,this paper also proposes a hybrid training method that combines the synchronous and asynchronous approaches;this significantly improves the training speed of the model without reducing the recognition accuracy.Experiments on the light source CXIDB76 public dataset reveal that FedKL can improve the accuracy and F1 score by 25.2 percentage points and 0.419,respectively,compared with FedAvg.

关 键 词：隐私保护 图像筛选 联邦学习 布拉格斑点 相对熵 

分 类 号：O432[机械工程—光学工程]