
Welcome!

This is a readme for the supplementary material for the Submodular + Concave paper.

File specific info is below.

	• The full_paper.pdf file is the complete paper along with the appendix (which contains most of the proofs along with some more experiment related plots).

The remaining files in the supplementary material are related to the code used in the experiments.

	• The interpolate.py file pertains to Section 4.1 in the main paper. After navigating to the directory where it is saved, just running the “python3 interpolate.py” command from the terminal should output a interpolate_fig.png image similar to Figure 2 in the main paper.

	• The quadraticprogramming.py file is related to Section 4.2.1, i.e. Figures 3a, 3b, 3c, or Figure 4 in Appendix B. In order to get any of the plots in these figures (say Figure 4a which is for n=12, m=6), open the quadraticprogramming.py file and enter the desired value of n (i.e. 12) and m (i.e. 6) in the designated part. By default both are set to 8. Then save and run “python3 quadraticprogramming.py” from the terminal. In the quadraticprogramming_fig.png file, you should have the desired plot, which is the average of 50 runs of the quadratic programming experiment corresponding to the set values of n and m.

	• The experimental_design.py file corresponds to Section 4.2.2, i.e. Figure 3d. Entering the desired dimension in the file and running “python3 experimental_design.py > experimental_design_output.txt” in the terminal will yield the final outputs of the various algorithms in the experimental_design_output.txt file. The default dimension is set to 8. The output is again the average of 50 runs. We run this script for n = 8, 12, and 16 and then plot the combined result in Figure 3d.

	• The jeopardy.py file corresponds to the experiment done on the Jeopardy dataset mentioned in Section 4.1 in the main paper. The code takes as input a data file (written put_your_input_data_here.txt in the code). The code is written for the questions.txt file which contains the Jeopardy questions. The questions.txt file is also included in the supplementary material. Running this code should output a output_file.txt which contains the similarity matrix (called L in the main paper) which can then be used with the code in interpolate.py.



