Monotone_Gradient_Networks

Maya Janvier and Margot Boyer

Implementation of Learning Gradients of Convex Functions with Monotone Gradient Networks (Chaudhari et al. (2023)) [1]

Reproducing the results

The main code is in the experiments.ipynb notebook. You will be able to reproduce the following experiments:

• Implementation of models: models.py and section 1 notebook
• Gradient field experiment from [1] : section 2 notebook
• Optimal coupling: section 3 notebook, 3.a Wasserstein loss, 3.b KL-divergence loss, 3.c CP-Flow experiment (see setup)
• Color domain adaptation: section 4 notebook

Setup

Running CP-Flow experiment (part 3.c)

• Clone CP-Flow to be able to run the corresponding experiment in the notebook
• Move models.py,train_ot_coupling.py into the main folder of CP-Flow
• Move into CP-Flow folder and run pip install -r requirements.txt
• Run python3 train_ot_coupling.py

Color adaptation experiment

If you want to run on more images, you can follow these steps:

• Download the Dark Zurich dataset validation set.
• Get the folder /Dark_Zurich_val_anon/rgb_anon/val_ref/day/GOPR0356_ref and add new pictures to the dark_zurich folder !