Skip to content

Commit

Permalink
Final tweaks to camera-ready version
Browse files Browse the repository at this point in the history
  • Loading branch information
mdbenito committed Jun 30, 2024
1 parent bdad896 commit a62ebba
Showing 1 changed file with 5 additions and 6 deletions.
11 changes: 5 additions & 6 deletions report/tmlr-re-csshapley.tex
Original file line number Diff line number Diff line change
Expand Up @@ -10,7 +10,7 @@
\usepackage{natbib}

\usepackage[bookmarks=true]{hyperref}
\usepackage{bookmark,hypcap, cleveref}
\usepackage{bookmark,hypcap,cleveref}
\hypersetup{
bookmarksnumbered=true,
bookmarksopen=true,
Expand Down Expand Up @@ -173,9 +173,9 @@ \section{Definitions and notation}
one can do for general $u$, since given any weighting scheme, it is always
possible to adversarially construct a utility with high variance at the set
sizes with highest weights. The authors do indeed prove certain optimality
results wrt. a notion of stability in value rankings. We will see in the
experiments that this simple idea yields the best results in many
situations.\footnote{{\citep{li_robust_2023}} recently extended DB to
results with respect to (wrt.) a notion of stability in value rankings.
We will see in the experiments that this simple idea yields the best results in
many situations.\footnote{{\citep{li_robust_2023}} recently extended DB to
{\tmem{weighted}} Banzhaf values, but we were not able to include this method
in our experiments.}

Expand Down Expand Up @@ -208,7 +208,7 @@ \section{Definitions and notation}
D\backslash D_{\tmmathbf{y}_i}$. Analogously, $T = T_{\tmmathbf{y}_i} \uplus
T_{-\tmmathbf{y}_i}$ is a partition into the subsets of all training data with
the same and different class than $\tmmathbf{y}_i$, respectively. Trained over
any $S \subseteq T$ the model has {\tmdfn{in-class accuracy}}: \footnote{We
any $S \subseteq T$ the model has {\tmdfn{in-class accuracy}}:\footnote{We
follow the notation of {\cite{schoch_csshapley_2022}}, but observe that the
sub-index in $a_S$ is a variable. A more obvious notation would be $a (S,
D_{\tmmathbf{y}_i})$.} $a_S (D_{\tmmathbf{y}_i}) \assign \left( \text{\#
Expand Down Expand Up @@ -3554,4 +3554,3 @@ \subsection{Value transfer}\label{app:value-transfer}
% IGdyYWRpZW50IGJvb3N0aW5nCiAgY2xhc3NpZmllciwgZm9yIHZhbHVlcyBjb21wdXRlZCB1c2luZyBL
% Tk4ufQpcZW5ke2ZpZ3VyZX17XGV0bXs3OH19CgpcZW5ke2RvY3VtZW50fQo=
%%%%%%%%%% End TeXmacs source

0 comments on commit a62ebba

Please sign in to comment.