\begin{tabular}{ll}
\toprule
{} &                                                                                                                                                                                          Citation \\
\midrule
Analysis of weights                                    &  \cite{Perez-Benitez2018, Yoon2018, Langkvist2018, Deiss2018, Lawhern2018, Xu2016, Tsinalis2016a, Nurse2016, Tabar2016a, Zheng2015, Stober2015, Manor2015, Yang2015a, Langkvist2012, Cecotti2011} \\
Analysis of activations                                &                                                                                           \cite{Yuan2018a, Waytowich2018, Lawhern2018, kwak2017, Yin2017a, Supratak2017, Shamwell2016, Manor2015} \\
Input-perturbation network-prediction correlation maps &                                                                                                              \cite{Schirrmeister2017a, Volker2018, Hartmann2018b, Behncke2017, Schirrmeister2017} \\
Generating input to maximize activation                &                                                                                                                                      \cite{VanPutten2018b, Ruffini2018a, Sors2018, Bashivan2016a} \\
Occlusion of input                                     &                                                                                                                                                        \cite{Lee2018, Chambon2018, Thodoroff2016} \\
Analysis of most-activating input windows              &                                                                                                                                                                              \cite{Hartmann2018b} \\
Retrieval of closest examples                          &                                                                                                                                                                                  \cite{Deiss2018} \\
Input-feature unit-output correlation maps             &                                                                                                                                                                          \cite{Schirrmeister2017} \\
Analysis of generated outputs                          &                                                                                                                                                                               \cite{Hartmann2018} \\
Analysis of performance with transferred layers        &                                                                                                                                                                            \cite{Hajinoroozi2017} \\
Saliency map                                           &                                                                                                                                                                               \cite{Vilamala2017} \\
Class activation maps                                  &                                                                                                                                                                                  \cite{Ghosh2018} \\
Ablation of filters                                    &                                                                                                                                                                                \cite{Lawhern2018} \\
Deeplift                                               &                                                                                                                                                                                \cite{Lawhern2018} \\
\bottomrule
\end{tabular}