treatment and outcome. The second is efficient language modeling: representations of text are designed to dispose of linguistically irrelevant information, and this information is also causally irrelevant. Our method adapts language models (specifically, word embeddings and topic models) to learn document...
Adapting Text Embeddings for Causal InferenceVictor VeitchDhanya SridharDavid M. BleiPMLRUncertainty in Artificial Intelligence
Causally sufficient embeddings combine two ideas. The first is supervised dimensionality reduction: causal adjustment requires only the aspects of text that are predictive of both the treatment and outcome. The second is efficient language modeling: representations of text are designed to dispose of ...
具体来说, \text{do} 符号表示对系统进行干预,例如,直接为一名新患者分配药物,并观察其结果。这样的因果效应与单纯的统计相关性有所不同,因为它强调了干预带来的变化。 如果观测到的协变量 X 包含了所有处理(treatment)和结果(outcome)的共同原因(即,阻断了所有“后门路径”),那么因果效应可以用观测分布 P 的某...