Skip to content
This repository has been archived by the owner on Jul 22, 2024. It is now read-only.

Latest commit

 

History

History

ERM-IRM

Empirical or Invariant Risk Minimization? A Sample Complexity Perspective?

Instructions on reproducing results from "Empirical or Invariant Risk Minimization? A Sample Complexity Perspective" link

Classification Experiments

  1. We have attached for Jupyter Notebooks a) Reproduce CS-CMNIST comparisons using IRM_CS_CMNIST_final.ipynb b) Reproduce CF-CMNIST comparisons using IRM_CF_CMNIST_final.ipynb c) Reproduce AC-CMNIST comparisons using IRM_AC_CMNIST_final.ipynb d) Reproduce HB-CMNIST comparisons using IRM_AC_CMNIST_final.ipynb

Regression Experiments

  1. Go to IRM_exhaustive_argparse_hper_wo_disc_Sep8.py and enter the path to the directory to save the plots in the already defined variable called dir_name
  2. Run IRM_exhaustive_wo_disc_Sep8.sh and it will generate the plots for the four regression setups (CS-Regression, CF-Regression,AC-Regression, HB-Regression)

Note our code for IRMv1 regression is based on original repository of IRM which has its own license