知识卡片

普通读书笔记卡 · 8836.c · 生成模型:GPT-6

噪声保护与统计精度有取舍

理解等级:中级

随机化能让个体回答更不确定,同时总体统计仍可经已知机制校正。更强扰动通常增加估计不确定性,需要更多样本或接受较低精度。保护声明要对应完整机制和参数,不能把汇总仍可用推断为个体信息绝不可知。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      it is still possible to infer the true percentage of “yes” answers in the surveyed population from the stored data: the expected value of the measured percentages will be the true “yes” and “no” answers in the population. The price for this differential privacy is that more data must be collected to obtain the same accuracy guarantees in the estimate of the true population averages.

    参考来源

    - 位置:《Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)》第04章《计算机视觉与社会》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Foundations of Computer Vision (Antonio Torralba, Phillip Isola etc.)/_epub-src/OPS/xhtml/c004.xhtml) - 结论依据:摘录中关于噪声保护与统计精度有取舍的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:it is still possible to infer the true percentage of “yes” answers in the surveyed population from the stored data: the expected value of the measured percentages will be the true “yes” and “no” answers in the population. The price for this differential privacy is that more data must be collected to obtain the same accuracy guarantees in the estimate of the true population averages.