/minimax_OT

Code for paper "A Swiss Army Knife for Minimax Optimal Transport"

Primary LanguageJupyter Notebook

minimax_OT

Code for paper "A Swiss Army Knife for Minimax Optimal Transport", ICML'20

Dependencies:

  • Numpy >= 1.18.1
  • POT >= 0.6.0
  • CVXPY >= 1.0.25
  • MOSEK >= 9.1.9
  • Umap >= 0.1.1
  • Scikit-Image >= 0.16.2
  • Scikit-Learn >= 0.22.1
  • Subspace Robust Wasserstein code

Running the experiments:

Main paper

Figure 2

  • Left: algoError.py
  • Middle: algoErrorVsReg.py
  • Right: algo_vs_LP.py

Figure 3

  • Left: fragmentedCube.py
  • Middle and right: Experiment.ipynb

Supplementary material:

Experiment.ipynb

For Mac users

execute Install Certificates.command in Python folder installation