pyRiemann-qiskit

Literature on quantum computing suggests it may offer an advantage compared with classical computing in terms of computational time and outcomes, such as for pattern recognition or when using limited training sets [1, 2].

A ubiquitous library on quantum computing is Qiskit [3]. Qiskit is an IBM library distributed under Apache 2.0 which provides both quantum algorithms and backends. A backend can be either your local machine or a remote machine, which one can emulate or be a quantum machine. Qiskit abstraction over the type of machine you want to use, makes designing quantum algorithms seamless.

Qiskit implements a quantum version of support vector-like classifiers, known as quantum-enhanced support vector classifiers (QSVCs) and variational quantum classifiers (VQCs) [4]. These classifiers likely offer an advantage over classical SVM in situations where the classification task is complex. Task complexity is raised by the encoding of the data into a quantum state, the number of available data, and the quality of the data. An initial study is available in [5], and it can be downloaded from here. Although there is no study on this topic at the time of writing, this could be an interesting research direction to investigate BCI illiteracy.

pyRiemann-qiskit implements a wrapper around QSVC and VQC, to use quantum classification with Riemannian geometry. A use case would be to use vectorized covariance matrices in the tangent space as an input for these classifiers, enabling a possible sandbox for researchers and engineers in the field.

pyRiemann-qiskit also introduces a quantum version of the famous MDM algorithm. There is a dedicated example on quantum-MDM here.

The remaining of this readme details some of the quantum drawbacks and will guide you through installation. Full documentation, including API description, is available at https://pyriemann-qiskit.readthedocs.io/. The repository also includes a wiki where you can find additional information.

Quantum drawbacks

• Limitation of the feature dimension

  The number of qubits (and therefore the feature dimension) is limited to:

  • ~36 (depends on system memory size) on a local quantum simulator, and up to:
  • 5000 on a remote quantum simulator;
  • 7 on free real quantum computers, and up to:
  • 127 on exploratory quantum computers (not available for public use).

• Time complexity

  A higher number of trials or dimensions increases the time to completion of the quantum algorithm, especially when running on a local machine. This is why the number of trials is limited in the examples we provided. However, you should avoid such practices in your own analysis.

  Although these aspects are less important in a remote backend, it may happen that the quantum algorithm is queued depending on the number of concurrent users.

  For all these aspects, the use of pyRiemann-qiskit should be limited to offline analysis only.

References

[1] A. Blance and M. Spannowsky, ‘Quantum machine learning for particle physics using a variational quantum classifier’, J. High Energ. Phys., vol. 2021, no. 2, p. 212, Feb. 2021, https://doi.org/10.1007/JHEP02(2021)212

[2] P. Rebentrost, M. Mohseni, and S. Lloyd, ‘Quantum Support Vector Machine for Big Data Classification’, Phys. Rev. Lett., vol. 113, no. 13, p. 130503, Sep. 2014, https://doi.org/10.1103/PhysRevLett.113.130503

[3] H. Abraham et al., Qiskit: An Open-source Framework for Quantum Computing. Zenodo, 2019, https://doi.org/10.5281/zenodo.2562110.

[4] V. Havlíček et al., ‘Supervised learning with quantum-enhanced feature spaces’, Nature, vol. 567, no. 7747, pp. 209–212, Mar. 2019, https://doi.org/10.1038/s41586-019-0980-2

[5] G. Cattan, A. Andreev, First steps to the classification of ERPs using quantum computation, NTB Berlin 2022 - International Forum on Neural Engineering & Brain Technologies, May 2022, Berlin, Germany, https://hal.archives-ouvertes.fr/hal-03672246/

How to cite?

Anton Andreev, Grégoire Cattan, Sylvain Chevallier, and Quentin Barthélemy. ‘pyRiemann-qiskit: A Sandbox for Quantum Classification Experiments with Riemannian Geometry’. Research Ideas and Outcomes 9 (20 March 2023). https://doi.org/10.3897/rio.9.e101006.

This library is part of the Qiskit Ecosystem

Installation

We recommend the use of Anaconda to manage python environements.

pyRiemann-qiskit currently supports Windows, Mac and Linux OS with Python 3.9 - 3.11.

You can install pyRiemann-qiskit release from PyPI:

pip install pyriemann-qiskit

The development version can be installed by cloning this repository and installing the package on your local machine using the setup.py script. We recommand to do it using pip:

pip install .

Note that the steps above need to be re-executed in your local environment after any changes inside your local copy of the pyriemann_qiskit folder, including pulling from remote.

To check the installation, open a python shell and type:

import pyriemann_qiskit

To enable Qiskit GPU optimization when using quantum simulation, run:

pip install .[optim]

Note, Qiskit only provide binaries for Linux. For other platforms, or if you want to enable specific NVIDIA optimization for quantum, you need to build the binary yourself.

To run a specific example on your local machine, you should install first dependencies for the documentation:

pip install .[docs]

Then you can run the python example of your choice like:

python examples\ERP\classify_P300_bi.py

Installation with docker

We also offer the possibility to set up the dev environment within docker. To this end, we recommand to use vscode with the Remote Containers extension from Microsoft.

Once the installation is successful, just open the project in vscode and enter F1. In the search bar that opens, type Rebuild and Reopen Container.

Wait for the container to build, and open a python shell within the container. Then ensure everything went smoothly by typing:

import pyriemann_qiskit

Alternatively, you can from the console (Windows or Linux) build the docker image from our Dockerfile. Go to the root folder of pyRiemann-qiskit and type:

docker build -t pyrq .

Next use docker run --detach pyrq to enter the pyRiemann-qiskit image.

If you wish, you can also download docker images directly from github docker registry: https://github.com/pyRiemann/pyRiemann-qiskit/pkgs/container/pyriemann-qiskit

They are pushed to the docker registry on each release.

Contributor Guidelines

Everyone is welcome to contribute to this repository. There are two types of contributions:

• Raise an issue on the repository. Issues can be either a bug report or an enhancement proposal. Note that it is necessary to register on GitHub before. There is no special template that is expected but, if you raise a defect please provide as many details as possible.

• Raise a pull request. Fork the repository and work on your own branch. Then raise a pull request with your branch against master. As much as possible, we ask you to:

  • avoid merging master into your branch. Always prefer git rebase.
  • always provide full documentation of public methods.

Code contribution (pull request) can be either on core functionalities, documentation or automation.

• The core functionalities are based on Python, pyRiemann, Qiskit ML and follow the best practice from scikit-learn. We use flake8 for code formatting. flake8 is installed with the testing dependencies (see below) or can be installed directly from pip:

  pip install flake8
  

  To execute flake8, just type flake8 from the root repository, and correct all errors related to your changes.

• The documentation is based on Sphinx.

• Automation is based on GitHub Action and pytest. It consists in two automated workflows for running the example and the tests. To run the tests on your local machine, you should first install the dependencies for testing:

  pip install .[tests]
  

  and then type pytest from the root repository. You can also specify a file like:

  pytest tests/test_classification.py
  

  Workflows are automatically triggered when you push a commit. However, the workflow for example execution is only triggered when you modify one of the examples or the documentation as the execution takes a lot of time. You can enable Github Actions in your fork to see the result of the CI pipeline. Results are also indicated at the end of your pull request when raised. However note, that workflows in the pull request need approval from the maintainers before being executed.

Troubleshooting

See our dedicated wiki page.