因果推断深度学习工具箱 - Learning Decomposed Representation for Counterfactual Inference
文章名称LearningDecomposedRepresentationforCounterfactualInference核心要点沿袭了DR-CFR的思路,把covariates分解成confounder,Adjustment和instrument,准确的找出confounder,解决因为控制非confounder而带来的bias。作者通过分析因果图,利用随机变量(条件)独立性,并最大化Adju