contributions.tex (935B)
1 \begin{frame}{Contributions} 2 \begin{block}{Regularizing Discriminative Methods} 3 \begin{itemize} 4 \item Trained a deep (PCNN) classifier. 5 \item Introduced two regularizing losses: 6 \begin{itemize} 7 \item A \alert{skewness} loss to ensure confidence. 8 \item A \alert{distribution distance} loss to ensure diversity. 9 \end{itemize} 10 \item Improved experimental setup: 11 \begin{itemize} 12 \item 2 metrics (V-measure, ARI).\quad\raise1.25pt\hbox{\textcolor{palette-1}{\(\bullet\)}} 2 datasets (T-RExes). 13 \end{itemize} 14 \end{itemize} 15 \end{block} 16 17 \begin{block}{Graph-based Aggregate Methods} 18 \begin{itemize} 19 \item Explicitly modeled the aggregate setup for the unsupervised problem. 20 \item Provided proof on the quality of topological information. 21 \item Proposed an approach to exploit the mutual information between topological and linguistic features. 22 \end{itemize} 23 \end{block} 24 \end{frame}