This repository contains PyTorch implementation of MD-GAN, along with training iPython notebook and trained models. Currently, this repository contains the training of data generated from a Gaussian mixture model (GMM). Two trained models included in this repository: the first one trained on data of a grid of 5 x 5 mixture of Gaussian and the second model trained on data of two mixture of Gaussian which are centered at -5 and 5.
The repository contains a training notebook for a Gaussian mixture model with 25 centers and a second notebook for Gaussian mixture model with two centers.
The trained model are located in this repository at the models folder.
| Data | Model Path | Training Notebook |
|---|---|---|
| 25 GMM | models/generator_25.pt | examples/training_md_gan_25_gaussions.ipynb |
| Two GMM | models/generator_2.pt | examples/training_md_gan_two_gaussions.ipynb |
Note: this code is not well documented yet, this will be done in the near future.
Example figures:
The left figure presents an example of real data and data generated using the MD-GAN D using the Gaussian mixture model with 25 and the right figure presents both the real data and generated data using MD-GAN while using the Gaussian mixture model with two centers.
If you find a bug or have a question, please create a GitHub issue.