Contents:

1. PYTHON: There are 5 python codes for training and testing the CNNs are prefixed by the phrase 'python_'.

2. (i)  MATLAB scripts/functions: There are 4 MATLAB scripts/function files (.m). 'main_BRP_test.m' is the main MATLAB script 
   and calls the 'fmincon_feasibility.m', 'fmincon_robust.m' and 'fmincon_sparse.m' functions.
   (ii) MATLAB data: There are 12 MATLAB data files (.mat) prefixed by the phrase 'MAT_'. The data files prefixed by 'MAT_dl_data_' contain the      deep image classification data aggregated by training the CNNs. The remaining '.mat' data files contain the sparsity enhanced as well
   as the robustness optimized representations of the interpretable deep image classification model for the CNNs. 

3. PDF file: There is 1 pdf file that includes the proof of Theorem 1, intuition behind Theorem 1, and additional numerical results obtained from Theorem 2.