Vanilla-GAN

---

This repository contains a simple implementation of Vanilla Generative Adverserial Network (paper) with the Generator and Discriminator being simple Multi-layer perceptrons. I also used the training tips-and-tricks from this paper and this paper. The model is trained on MNIST digit dataset.

Loss curves (blue: generator; orange: discriminator)

Generated samples

How to run the code

Just execute

python gan.py

Requirements:

1. PyTorch
2. Matplotlib
3. Numpy