Quantum GAN – Variational Quantum Generator + Classical Discriminator

Quantum GAN: PennyLane + PyTorch

This project implements a Quantum Generative Adversarial Network (QGAN) combining a variational quantum generator with a classical discriminator, for 2D synthetic data generation and visualization.

Project Description

This project implements a Quantum Generative Adversarial Network (QGAN) with:

• Quantum Generator: using Variational Quantum Circuit (VQC) via PennyLane
• Classical Discriminator: Multi-Layer Perceptron (MLP) implemented in PyTorch

Project Goals:

1. Explore the combination of quantum computing and classical machine learning.
2. Generate 2D synthetic data in the shape of a noisy circle resembling real data.
3. Provide visualization and training loss analysis.
4. Demonstrate potential of Quantum GANs for physics simulations, biological data, and synthetic datasets.

The project also supports Qiskit Aer, enabling quantum expectation inference and noise simulation.

---

Key Features

• Quantum generator using PennyLane QNode
• Simple yet effective classical discriminator
• Synthetic 2D circular noisy dataset
• Visualization:
  • Real vs Generated data
  • Generator & Discriminator loss curves
• Optional support for Qiskit Aer noise simulation

---

---

Installation

Recommended to use Python 3.12 and a virtual environment:

# 1. Install Python 3.12 via Homebrew
brew install python@3.12

# 2. Create virtual environment
python3.12 -m venv venv
source venv/bin/activate

# 3. Upgrade pip
pip install --upgrade pip

# 4. Install dependencies
pip install numpy matplotlib pennylane torch qiskit qiskit-aer

# Activate virtual environment
source venv/bin/activate

# Run the QGAN
python3 main.py

results/qgan_training_summary.png
😊🫰