增强深度学习中的差分隐私防御机制

余方超, 方贤进, 张又文, 杨高明, 王丽

Enhanced differential privacy defense mechanism in deep learning

Fangchao Yu, Xianjin Fang, Youwen Zhang, Gaoming Yang, Li Wang

图11 η=0.05时的测试集损失（a）和预测准确率（b）

Fig.11 Testing set loss (a) and prediction accuracy (b) when η=0.05