GANs

Objectives

The objectives of this project is to develop a quantum GAN using parameterized quantum circuits in order to be able to compare both classical and quantum models and study training stability, sample quality or convergence behavior.

This project is part of a broader exploration of Quantum Computing and Quantum Machine Learning.

Background

Generative Adversarial Networks consist of two competing models:

• Generator (G): Generates synthetic data
• Discriminator (D): Distinguishes real data from generated data

The objective function is describe as follows:

$$\min_{G} \max_{D} V(D, G) = \mathbb{E}_{x \sim p_{data}} [\log D(x)] + \mathbb{E}_{z \sim p_{z}} [\log(1 - D(G(z)))]$$

One objective of G is to minimize V, whereas another objective of G is to maximize V.

CycleGANs consist of two generators and two discriminators. The generators learn to translate images from domain X to domain Y and vice versa, while the discriminators aim to distinguish between real and generated images in both domains. We'll define the architecture for both the generator and the discriminator, using a standard architecture for CycleGAN.

Structure

.
├── data/
│   └── dataset.py
│
├── qgan/              
│   ├── generator.py
│   ├── discriminator.py
│   └── vqc.py
│
├── experiments/
├── main.ipynb
├── LICENSE
└── README.md

Model performance

Author

Project build by Antony Manuel and inspired by Omar Ikne's courses, PhD student at IMT Nord Europe.