diff --git a/paper/aaai/paper.pdf b/paper/aaai/paper.pdf index c7b1d817cfc62c1a20a07ac0a42ab623b052c343..d7b54fce2f6aacbeadf6dd4f44a587a2bccf4ca6 100644 Binary files a/paper/aaai/paper.pdf and b/paper/aaai/paper.pdf differ diff --git a/paper/appendix.tex b/paper/appendix.tex index 5ef3a48b07b2b5217b53aaf5ceacefffa9bc0a07..d402dda38016081bb1f65ba50fe7c8ec8a03bd05 100644 --- a/paper/appendix.tex +++ b/paper/appendix.tex @@ -120,8 +120,7 @@ Aside from the additional penalties in Equation~\ref{eq:eccco-app}, the only key \end{aligned} \end{equation} -we take these parameters as fixed in our counterfactual search. - +During the counterfactual search we take the network parameters as fixed and instead optimize with respect to counterfactual itself: \subsubsection{A Note on Convergence} diff --git a/paper/body.tex b/paper/body.tex index 34ef9e2b6bce0eb92dfde1c4a6f9e4d25829da68..47bf65eea36d73ea6405a6bc3c1b331b05866a8f 100644 --- a/paper/body.tex +++ b/paper/body.tex @@ -142,7 +142,7 @@ Equation~\ref{eq:impl} gives rise to a similar evaluation metric for unfaithfuln \end{aligned} \end{equation} -Our default choice for the $\text{dist}(\cdot)$ in both cases is the L1 Norm, which is consistent with the common choice for evaluating counterfactuals with respect to the closeness desideratum~\citep{pawelczyk2021carla,pawelczyk2022probabilistically}. +Our default choice for the $\text{dist}(\cdot)$ function in both cases is the Euclidean Norm. \section{Energy-Constrained Conformal Counterfactuals}\label{meth}