Quantum Control with Neural Networks: PINN and QubiNN

This repository contains the Jupyter notebooks, training data, and saved models used in the project "Quantum Control with Neural Networks: PINN and QubiNN", which explores the application of neural networks for single-qubit quantum control tasks.

Project Overview

The project investigates two neural network architectures:

• Physics-Informed Neural Networks (PINNs): Incorporates the Schrödinger equation to enforce physical consistency in predicting control fields and expectation values.
• QubiNN: A purely data-driven feedforward neural network used as a baseline for comparison.

Both models were trained and benchmarked using data generated via QuTiP simulations. The results highlight the trade-offs between accuracy, physical consistency, and computational efficiency.

---

Repository Structure

Folders

• notebooks/: Contains the main Jupyter notebooks used in the project:

  • training_data.ipynb: Generates training data using QuTiP simulations.
  • QuPINN.ipynb: Implements and trains the PINN for predicting control fields and expectation values.
  • QubiNN.ipynb: Implements and trains the QubiNN for comparison.
  • benchmarking.ipynb: Benchmarks the performance of PINN and QubiNN against QuTiP simulations.

• data/: Contains the training data used for the models, including:

  • training_data.csv: The dataset of control fields and expectation values generated from QuTiP simulations.

• models/: Contains the saved model weights for PINN and QubiNN:

  • QuPINN_model.pth
  • QubiNN_model.pth

Additional Files

• README.md: This documentation file.
• requirements.txt: A list of dependencies needed to run the notebooks.

---

How to Use

1. Clone the Repository

git clone https://github.com/your-username/quantum-control-neural-networks.git
cd quantum-control-neural-networks

2. Set Up the Environment

Create and activate a Python virtual environment, then install the required dependencies:

pip install -r requirements.txt

3. Run the Notebooks

Navigate to the notebooks/ folder and open the desired notebook using Jupyter Notebook or JupyterLab:

jupyter notebook notebooks/training_data.ipynb

---

Key Features

1. Training Data Generation:
   • The training_data.ipynb notebook uses QuTiP to generate control fields and expectation values for diverse target states on the Bloch sphere.
2. Model Training:
   • The QuPINN.ipynb and QubiNN.ipynb notebooks train the respective models on the generated data.
3. Benchmarking:
   • The benchmarking.ipynb notebook evaluates the models on test target states, comparing their performance against QuTiP simulations in terms of accuracy, fidelity, and computational efficiency.

---

Results Summary

• Performance:
  • QubiNN excels in predicting expectation values and achieves higher computational efficiency.
  • PINN slightly outperforms in control field predictions but is more computationally intensive.
• Speedup:
  • Both models achieve significant speedup compared to QuTiP simulations, with QubiNN achieving up to ~5000x faster predictions.

For detailed results and visualizations, refer to the benchmarking.ipynb notebook.

---

Dependencies

The project requires the following Python libraries:

• numpy
• matplotlib
• torch
• qutip
• pandas
• scikit-learn

These can be installed using the requirements.txt file.

---

Contact

For questions, suggestions, or collaboration inquiries, feel free to contact:

• Name: Diego Alducin
• Email: diego.alducin@gmail.com
• GitHub: datdiego