conclusion.tex (626B)
1 \begin{frame}{Conclusion} 2 \begin{block}{Take-home Message} 3 Selecting good regularizations to enforce modeling hypotheses enables us to train a deep classifier. 4 \end{block} 5 \begin{block}{Contributions} 6 \begin{itemize} 7 \item Train a PCNN without supervision 8 \item Designed two regularization losses (Skewness, Distribution distance) 9 \item Introduced new datasets (T-RExes) 10 \item Evaluated using additional metrics (V-measure, ARI) 11 \end{itemize} 12 \end{block} 13 14 \medskip 15 16 Étienne Simon, Vincent Guigue, Benjamin Piwowarski. {\futuraHeavier \citefield{fitb}[linkedtitle]{title}} ACL~\cite*{fitb} 17 \end{frame}