• Lasso
  • Risk of lasso and optimal lasso ensemble (Figures 4, 5 and 12):
    • run_lasso_opt.py
  • Risk of full lasso ensemble (Figures 6 and 14):
    • run_lasso_equiv.py
  • Risk of optimal lasso ensemble (Figure 7):
    • run_lasso_opt_2.py
  • Fixed-point quantities of lassoless (Figure 10):
    • run_lassoless.py
  • Empirical risk of lassoless ensemble (Figures 11, 13):
    • run_lasso_emp.py
  • Risk of optimal lasso ensemble with anisotropic covariance (Figure 9)
    • run_lasso_opt_aniso.py
• Huber
  • Risk of full unregularized Huber ensemble (Figure 15):
    • run_huber.py
  • Risk of l1-regularized Huber and optimal l1-regularized Huber ensemble (Figures 3):
    • run_huber_l1_opt.py
  • Risk of full l1-regularized Huber ensemble (Figures 2, 16 and 17):
    • run_huber_l1_emp.py
    • run_huber_l1_equiv.py
  • Numerical evaluation of theoretical quantities with anisotropic covariance (Figure 8)
    • ell1_huber_aniso.ipynb
• Utility functions
  • compute_risk.py
  • generate_data.py
• Visualization
  • The figures can be reproduced with the Jupyter Notebook Plot.ipynb.

All the experiments are run on Ubuntu 22.04.4 LTS using 12 cores and 128 GB of RAM.

The estimated time to run all experiments is roughly 12 hours.