Introduction

This repository uses a Recurrent Neural Network (with 3-stacked GRU cells) for reconstructs quantum states/

This implementation is based on the findings of the following paper .

Installation Requirements

Before running the code, please ensure you have the following:

• Python 3.0
• Pytorch v0.4.1

Getting Started

For a quick start, please ensure the following.

• Clone the repository:

  In an appropriate directory run the following command on your Terminal:

  git clone https://github.com/akshat998/3-GRU-Phases-of-Matter.git

• Make sure you cd into the right directory.

  cd 3-GRU-Phases-of-Matter/

• Create your data

  By following the above instructions, the data will be saved in a folder of directory:

  /data_creator/DATA

  Please ensure you remember the name of your data directory.

• Train your model:

  The Recurrent Neural Network has been created using PyTorch. To start training, please run the following command on your Terminal:

  python3 RNN_Torch.py

  You will be asked to enter some parameters prior to training.

  After each batch, an error is printed & the model is saved in the directory (with format EnteredName_batchNumber):

  ./saved_models

• Evaluate your model:

  After training your model, you can test the quality of your model (i.e. calculate Classical & Quantum Fidelity). To do so, please run:

  python3 sampling.py

  Upon running, the user will be asked to provide:

  • Name of the saved model
  • Number of samples to be generated from the saved model
  • Name of the original data file that was used to train the saved mode

  Please note: For smaller systems, Quantum Fidelity is calculated using the overlap matrix. And for larger systems, Quantum Fidelity is calculated using a Monte-Carlo Approximation

Questions, problems?

Make a github issue 😄. Please be as clear and descriptive as possible. Please feel free to reach out in person: (akshat[DOT]nigam[AT]mail[DOT]utoronto[DOT]ca)

Acknowledgment

• A special thanks to the author of the Reconstructing quantum states with generative models: Dr. Juan Carrasquilla

• Also, the creator of ncon.py: __________