知识卡片
噪声重命名未必改变反事实
把对称噪声N全局换名为−N,若所有事实推断及干预都相应一致变换,模型可能在目标反事实上仍等价。不能仅以相同数字n代入后输出不同,就证明两个模型的反事实预测一定不同;应比较完整跨情境对应。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
- · 结构因果模型
原书出处 · 1 处
- · 查看原书依据
subtracts ny instead of adding ny. A normal distribution is symmetric around its mean, so since ny has a normal distribution with mean 0, the probability values of ny and –ny are the same, so the probability distribution of Y is the same in both models. But for the same values of ny and x, the actual assigned values of y will be different. Next,
参考来源
- 位置:《Causal AI》第06章《结构因果模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-6.xhtml)
- 结论依据:摘录中关于噪声重命名未必改变反事实的定义、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:subtracts ny instead of adding ny. A normal distribution is symmetric around its mean, so since ny has a normal distribution with mean 0, the probability values of ny and –ny are the same, so the probability distribution of Y is the same in both models. But for the same values of ny and x, the actual assigned values of y will be different. Next,