人工智能培训

搜索

深度学习论文:用于无监督学习的差分专用混合类型数据生成(Differentially Private Mixed-Type Data Generation Fo

[复制链接]
1yuqunchenggong 发表于 2019-12-9 13:06:44 | 显示全部楼层 |阅读模式
1yuqunchenggong 2019-12-9 13:06:44 223 0 显示全部楼层
深度学习论文:用于无监督学习的差分专用混合类型数据生成(Differentially Private Mixed-Type Data Generation For Unsupervised  Learning)在这项工作中,我们介绍了用于合成数据生成的DP-auto-GAN框架,该框架结合了自动编码器的低维表示形式和生成对抗网络(GAN)的灵活性。该框架可用于接收原始敏感数据,并私下训练一个模型,以生成将满足与原始数据相同的统计特性的合成数据。该学习的模型可用于生成任意数量的公共合成数据,由于后处理保证了差异性隐私,因此可以自由共享这些数据。我们的框架适用于未标记的混合类型数据,其中可能包括二进制,分类和实值数据。我们在未标记的二进制数据(MIMIC-III)和未标记的混合类型数据(ADULT)上都实现了此框架。我们还引入了新指标来评估合成混合类型数据的质量,尤其是在无人监督的情况下。
In this work we introduce the DP-auto-GAN framework for synthetic datageneration, which combines the low dimensional representation of autoencoderswith the flexibility of Generative Adversarial Networks (GANs).This frameworkcan be used to take in raw sensitive data, and privately train a model forgenerating synthetic data that will satisfy the same statistical properties asthe original data.This learned model can be used to generate arbitrary amountsof publicly available synthetic data, which can then be freely shared due tothe post-processing guarantees of differential privacy.Our framework isapplicable to unlabeled mixed-type data, that may include binary, categorical,and real-valued data.We implement this framework on both unlabeled binary data(MIMIC-III) and unlabeled mixed-type data (ADULT).We also introduce newmetrics for evaluating the quality of synthetic mixed-type data, particularlyin unsupervised settings.深度学习论文:用于无监督学习的差分专用混合类型数据生成(Differentially Private Mixed-Type Data Generation For Unsupervised  Learning)
URL地址:https://arxiv.org/abs/1912.03250     ----pdf下载地址:https://arxiv.org/pdf/1912.03250    ----深度学习论文:用于无监督学习的差分专用混合类型数据生成(Differentially Private Mixed-Type Data Generation For Unsupervised  Learning)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

1yuqunchenggong当前离线
新手上路

查看:223 | 回复:0

快速回复 返回顶部 返回列表