MNISQ: A Large-Scale Quantum Circuit Dataset for Machine Learning on/for Quantum Computers in the NISQ era

Welcome to MNISQ, a powerful resource designed to propel Quantum Machine Learning forward during the NISQ era. As we undergo the review process, anticipate exciting additions in the near future!

NEXT RELEASES

Tutorials for running the dataset with noise models!

Accessing the Datasets

Our datasets are now fully accessible, offering a plethora of opportunities for exploration and experimentation.

Executing with Qiskit

To leverage the dataset using Qiskit, refer to our comprehensive guide: Qiskit Quickstart Notebook.

Executing with PennyLane

For executing the dataset using PennyLane, we've prepared a dedicated guide: PennyLane Tutorial Notebook.

Local Dataset Download

Should you prefer working with datasets locally and employing the standard QASM formalism, our guide Downloading Datasets will walk you through the process.

Getting Started

Embark on your journey with the MNISQ dataset and Quantum Machine Learning experiments.

Join us in pushing the boundaries of Quantum Machine Learning during the NISQ era! Your exploration starts here.

This library provides machine learning datasets as a Qulacs's QuantumCircuit instance.

How To Install

pypi

pip install mnisq

source install

pip install git+https://github.com/FujiiLabCollaboration/MNISQ-quantum-circuit-dataset.git

Directory structure

• doc/source/notebooks provides sample usage of this library.
• mnisq and tests contains python code.
• generator_scripts contains scripts used to generate datasets.

Dataset accessibility

Data Construction

The datasets we created are easily accessible through the MNISQ library (currently in preparation). To install the library from PyPI, use the command: pip install mnisq.

The library automatically downloads predefined quantum circuits for Qulacs (Quantum Circuit Simulator), which are ready for use. When executing a circuit starting from the initial zero state, one of the images from the dataset is embedded in the quantum state.

Note: The QASM files come in two different forms:

• QASM with Dense() formalism: These files can be used in qulacs, but not on qiskit or other platforms due to the proprietary Dense() operator.
• Base QASM formalism: These files are compatible with any platform, including qiskit. You can find a tutorial on how to run them on our GitHub page at the following link: tutorial link. Please note that to access these files, they begin with the prefix "base_".

The datasets can be found at the following URL:

https://qulacs-quantum-datasets.s3.us-west-1.amazonaws.com/[data]_[type]_[fidelity].zip

Here are the parameters for the URL:

• data:

  1. "train_orig" (60,000 original encoded training data, QASM files with Dense() formalism).
  2. "base_train_orig" (same as above but the QASM files do not include Dense() operator, but a gate conversion. Can be run on qiskit and other platforms).
  3. "train": "train_orig" but augmented to 480,000 training data for each subdataset. Dense() formalism.
  4. "test": 10,000 test element from original encoding. QASM files with Dense() formalism.
  5. "base_test": same as above but the QASM files do not include Dense() operator, but a gate conversion. Can be run on qiskit and other platforms.

• type:

  1. "mnist_784": MNIST dataset.
  2. "Fashion-MNIST"
  3. "Kuzushiji-MNIST"

• fidelity:

  1. "f80": fidelity greater than or equal to 80%.
  2. "f90": fidelity greater than or equal to 90%.
  3. "f95": fidelity greater than or equal to 95%.

Example: https://qulacs-quantum-datasets.s3.us-west-1.amazonaws.com/test_mnist_784_f90.zip

Contributors

This project was developed by:

• Koki Aoyama(@kotamanegi)
• Hayata Morisaki
• Kouki Kawamura(@KowerKoint)
• Toshio Mori(@forest1040)
• Leonardo Placidi(@Gruntrexpewrus)
• Ryuichiro Hataya
• Kosuke Mitarai
• Keisuke Fujii

Citation

If you find the MNISQ dataset valuable for your research or work, please consider citing our paper:

@misc{placidi2023mnisq,
      title={MNISQ: A Large-Scale Quantum Circuit Dataset for Machine Learning on/for Quantum Computers in the NISQ era}, 
      author={Leonardo Placidi and Ryuichiro Hataya and Toshio Mori and Koki Aoyama and Hayata Morisaki and Kosuke Mitarai and Keisuke Fujii},
      year={2023},
      eprint={2306.16627},
      archivePrefix={arXiv},
      primaryClass={quant-ph}
}

Your acknowledgment helps us in further advancing the field of Quantum Machine Learning and fostering a collaborative research community.