This repository contains the source code for the paper: Learning hard distributions with quantum-enhanced Variational Autoencoders

• The sourcecode is available under the folder Sourcecode with sample notebooks in Notebooks.
• All datasets used for the experiments are present in Datasets. These include 4-qubit and 8-qubit states obtained from product states, haar states, quantum-circuit states, and quantum-kicked rotor states.

1. Clone this Github repository using the following command in your command line/terminal :
   git clone [email protected]:Anantha-Rao12/Quantum-enhanced-variational-autoencoder.git

2. Create a Python (>=3.2) virtual environemnt and call it 'Decoding-Quantum-States-with-NMR-env'.

• On Linux/ MacOS : python3 -m venv qiskitml-env
• On Windows : python -m venv qiskitml-env

A new directory called qiskitml-env will be created.

3. Activate the Virtual Environment by running:

• On Linux/ MacOS: source qiskitml-env/bin/activate
• On Windows: .\qiskitml-env\Scripts\activate

4. In the new virtual environemnt , run pip3 install -r requirements.txt to install all dependencies. On Windows, pip3 should be replaced by pip.

You are ready to start experiemnting with the code!

For the sake of simplicity and quick execution, we have a QeVAEWrapper() class that implements a QeVAE with default parameters. The model can be initialized, trained, and samples can be generated as shown in the notebook How to Train a QeVAE.ipynb. The three main steps are:

1. Creating the dataloaders : dataloaders, dataloader_info = setup_dataloaders(dataset, want_datasetsize=1, train_size=0.75)

2. Creating and fitting the mode:

qevae = QeVAEWrapper(num_qubits=2, latentsize=1)
qevae.fit(traindataloader, validdataloader, original_results=dataset_dict)

3. Finally, after training, we can generate samples using qevae.sample()

Under the Notebooks directory, we have two notebooks that show how we train QeVAEs for two tasks: (1) Learning the distribution (2) Compressing the circuit. The respective notebooks are titled 4qubit_productstates_analysis.ipynb and Circuit_compilation_QeVAE.ipynb contain more details on implementation.

DM Anantha S Rao - @anantharao00
For clarifications and queries -- Anantha Rao @2023

Project Link: https://github.com/AnanthaRao-12/Quantum-enhanced-variational-autoencoder