Skip to content

Latest commit

 

History

History
15 lines (12 loc) · 965 Bytes

README.md

File metadata and controls

15 lines (12 loc) · 965 Bytes
Raw

decongestion_toy_examples_final

Code for training critics to distinguish between 2-D distributions using Minibatch Optimal Ray Selection (MORS). Results from this code appear in my PhD thesis.

How to run this code

  • Create a Python virtual environment with Python 3.8 installed.
  • Install the necessary Python packages listed in the requirements.txt file (this can be done through pip install -r /path/to/requirements.txt).
  • Run critic_trainer.py with a chosen source and target distribution.

Important arguments for critic_trainer.py

  • source and target; which type distribution to use for $\mu$ and $\nu$. Several options are available.
  • source_params and target_params; further specifies the distributions $\mu$ and $\nu$. Enter 0 to see the syntax for your chosen type
  • save_dir; where the resulting figures will be saved
  • ot; if True, uses Minibatch Optimal Ray Sampling (MORS)
  • p; determines the cost function for MORS. $c(x,y) = |x-y|^p$