Supplementary Material Entropic Causal Inference: Identifiability and Finite Sample Results A Related Work

This is called the additive noise model (ANM) . This renders methods which heavily rely on the variable values, such as ANMs, unusable. We show that H ( X | Y = 2) is large using the above bound. Uniformity follows from uniformity of f . Applying Lemmas 3-5, with some derivation we show in Section F of the supplement that: Proposition 1 ( Step 6).