知识卡片
机制独立形成半监督方向假说
在简单且无隐藏混杂、输入分布与因果机制独立等设定下,因果方向的P(X)与P(Y|X)可分开,反因果分解则可能让无标签X对预测有用。这是带条件的研究假说,不是所有半监督算法的普遍不可能定理,更不保证反因果任务必然获益。
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
原书出处 · 1 处
- · 查看原书依据
only in the anti-causal case. See the chapter notes at www.altdeep.ai/causalAIbook for a more detailed explanation and references. But intuitively, we can see that this mirrors the the sunscreen and sunburn example—knowing the prevalence of sunburns P (O ) helped in learning how to guess sunscreen use when you know if someone has a sunburn P (C |O ). In this same anti-causal learning case,
参考来源
- 位置:《Causal AI》第05章《连接因果与深度学习》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-5.xhtml)
- 结论依据:摘录中关于机制独立形成半监督方向假说的定义、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:only in the anti-causal case. See the chapter notes at www.altdeep.ai/causalAIbook for a more detailed explanation and references. But intuitively, we can see that this mirrors the the sunscreen and sunburn example—knowing the prevalence of sunburns P (O ) helped in learning how to guess sunscreen use when you know if someone has a sunburn P (C |O ). In this same anti-causal learning case,
- 外部限定依据:[Schölkopf等,On Causal and Anticausal Learning(2012)](https://icml.cc/2012/papers/625.pdf),引言及结论明确简单无隐藏混杂设定、独立机制假说与初步元分析;非所有半监督学习的不可能定理。