Improvements of ML benchmark outputs

• Adjustment on the figures
• Add new figures
• Add a new method of preparing event data (imaging)

benchmarks/imaging_shower_ML/scripts/epcut_scan.py

@@ -231,7 +231,7 @@ if __name__ == '__main__':
                 # nsamples=npar, # this is redundant
                 nsamples_after_cut=ncut,
                 )
-            ax.hist(ep_centers, weights=vals, bins=50, label='${}$'.format(par.latex),
+            ax.hist(ep_centers, weights=vals/npar, bins=50, label='${}$'.format(par.latex),
                     color=colors[k], ec=colors[k], alpha=0.5)
             texts.append(r'$\epsilon_{{{}}}={:.2f}$%'.format(par.latex, eff*100.))
         cut_data.append(cut_result)
@@ -239,7 +239,7 @@ if __name__ == '__main__':
         ax.legend(fontsize=20, ncol=2, loc='upper center', bbox_to_anchor=(0.5, 1.12),)
         ax.tick_params(labelsize=20)
         ax.set_xlabel('$E/p$', fontsize=20)
-        ax.set_ylabel('Counts', fontsize=20)
+        ax.set_ylabel('Normalized Counts', fontsize=20)
         ax.axvline(x=ep_cut, color='k', ls='--', lw=2)
         ax.text(0.5, 0.97, '\n'.join(texts),
                 transform=ax.transAxes, fontsize=20, va='top', ha='center', bbox=box_props)