知识卡片
双重学习不只训练两个模型
DML先估计处理与结果的干扰函数,再以残差和相应正交矩条件估计处理效应;交叉拟合让每个单位的干扰预测来自其他折,减轻过拟合影响。正交性不消除未测混杂,也不保证任意学习器与任意收敛速度下推断有效。
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
Double machine learning Double machine learning (double ML) is a backdoor estimator that uses machine learning methods to fit two predictive models: a model of the outcome, given the adjustment set of confounders, and a model of the treatment, given the adjustment set. The approach then combines these two predictive models in a final-stage estimation to create a model of the target causal effect query.
参考来源
- 位置:《Causal AI》第11章《构建因果推断工作流》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-11.xhtml)
- 结论依据:摘录中关于双重学习不只训练两个模型的定义、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:Double machine learning Double machine learning (double ML) is a backdoor estimator that uses machine learning methods to fit two predictive models: a model of the outcome, given the adjustment set of confounders, and a model of the treatment, given the adjustment set. The approach then combines these two predictive models in a final-stage estimation to create a model of the target causal effect query.