# Paper: Decreasing Entropic Regularization Averaged Gradient for Semi-Discrete Optimal Transport

This repository contains the code for the paper **"Decreasing Entropic Regularization Averaged Gradient for Semi-Discrete Optimal Transport"**, submitted to NeurIPS 2025.

## 🧪 Environment Setup

To reproduce our experiments, you can recreate the exact environment using Conda. 
Then, run:

```bash
conda env create -f environment.yml
conda activate DRAG


## 📝 Explanations

Each experiment from the paper has its own dedicated notebook:

- **Synthetic data (convergence rate)**: can be run on a CPU and reproduces the theoretical convergence results.
- **MK-Quantiles**: requires a gpu
- **Generative model task**: located in the `Generative_Task/` folder and requires a GPU. 
