Welcome to the Quantum Generative Adversarial Network (QGAN)! This project explores the fusion of quantum computing and generative adversarial networks (GANs) to push the boundaries of machine learning and quantum technology.
QGAN combines the powerful framework of GANs with quantum circuits to generate data in quantum-enhanced ways. Leveraging PennyLane and PyTorch, this implementation enables quantum-inspired generators and discriminators, promising potential breakthroughs in various domains such as quantum chemistry, finance, and data generation.
- Quantum Discriminator & Generator: Powered by quantum circuits using the PennyLane library.
- Customizable Quantum Layers: Configure the number of qubits, quantum gates, and layers in the quantum circuits.
- Fully Integrated Training Pipeline: A complete pipeline to train the QGAN models with ease.
- Quantum Chemistry Support: Quantum simulations with molecular geometry support.
- Pythonic: Built with Python, using PyTorch for optimization and training.
-
Clone this repository:
git clone https://github.com/QuantaScriptor/QGAN_Project.git cd QGAN_Project-main -
Install the required dependencies:
pip install -r requirements.txt
-
Run the deployment script to start training:
bash deployment/deploy.sh
After installing the dependencies, you can modify the train.py script or directly run it for training your QGAN model. The key commands:
- Training the QGAN model:
python qgan/train.py
- Interacting with molecular problems: You can simulate molecules using the built-in
Molecule()function to create molecular geometries.from qgan.utils import Molecule molecule = Molecule("H2O") # Simulates H2O molecule
This project introduces a unique quantum twist on GANs:
- DiscriminatorQuantumCircuit: A quantum circuit acting as the discriminator, capable of detecting real vs. fake quantum states.
- GeneratorQuantumCircuit: A quantum circuit that generates synthetic data based on a latent quantum state.
By leveraging quantum gates, qubits, and quantum entanglement, the QGAN may offer faster convergence, better data generation, and possibilities for applications in quantum chemistry, quantum finance, and data science.
- GANs (Generative Adversarial Networks): GANs are a class of machine learning frameworks where two neural networks contest with each other in a game. In QGAN, these are replaced by quantum circuits.
- Quantum Circuits: At the heart of QGAN, quantum circuits encode data and perform quantum operations to simulate and compute outputs in ways classical computers cannot.
- Quantum Chemistry: You can define molecular structures and perform simulations using
pennylane_qchem.
- Generator creates quantum states that resemble real data.
- Discriminator evaluates if the generated data is real or fake.
- The system iterates until the generator produces quantum data indistinguishable from real data.
- Python 3.7+
- PennyLane and PyTorch
- Optional: Quantum simulators (e.g., IBM Q, Rigetti) for running on real quantum devices
This project is licensed under a Commercial Use License and Public Non-Commercial License. See the LICENSE file for more details.
This project was developed by Reece Colton Dixon and utilizes the amazing tools from PennyLane and PyTorch.
🌐 Quantum Generative Adversarial Networks (QGANs) offer a glimpse into the future of quantum machine learning. Explore the world of quantum data generation today!