
'empirical_error.py' does the main experiment, 'plot_empirical_errors.py'
produces the plots in the paper (Figure 2 and 3) given the outputs of
this experiment.
'num_parameters_plot.py' produces Figure 4.

'networks.py' implements in idiomatic pytorch the algorithm
described in Appendix E to write a general R-estimator as a
pytorch sequential of alternating linear and relu layers.
In particular 'build_shallowest_network' is a nice entry point.

Some paths will need to be set in an obvious way.

Gurobi is required, an academic license is available at
https://www.gurobi.com/downloads/end-user-license-agreement-academic/

