mezze

A toolbox for simulating open quantum system dynamics

Getting Started

For users who just want to use the master equation solver, channel classes, and/or cirq-based SchWARMA simulation cloning the repo and running pip install . from the project directory should work fine. The install can be verified by installing nose and running nosetests from the project directory. Some tests may fail with extremely small numerical differences based on random inputs.

The basic install does not include the faster Tensorflow-based simulation routines as well as the inference routines. Note that some of the example notebooks use this capability. To use these capabilities, see the next session.

Advanced Install -- Tensorflow-based simulation and inference

There are some issues with dependencies that may occur when using pip only so instead follow these steps:

• Install TensorFlow Quantum using the instructions here
  • Note sometimes there are issues with these directions, you may need to try other versions of tensorflow and tensorflow-quantum (in particular tfq-nightly will often work)
  • If installation fails on sympy, clone and install via git or with conda install sympy and repeat above instructions
• Install mezze - pip install . (or similar)

Even if there are pip issues with dependencies, as long as the tests run you should be fine.

Usage

Look at the notebooks in the examples folder, and also tests for more granular usage. The example notebooks have the following structure:

• examples - notebooks in this directory cover the older mezze functionality for manipulating quanutum states and channels, as well as stochastic master equation simulation
• examples/SchWARMAPaperExamples - notebooks in this directory generate the figures for Ref. [1], and are instructive for the finer details of SchWARMA, but most of that functionality has incorporated in a faster and more flexible codebase in mezze.mtfq
• examples/TFQ - notebooks that cover the new functionality uses cirq objects and "SchWARMAFies" them to induce spatiotemporally correlated noise, these can then be simulated by simulators from cirq, tensorflow-quantum, and qsim (assuming the latter two are installed). This directory also contains an example of how to do noise injection in the IBMQE as in [2]. The mtfq codebase also supports heavy-tailed distributions as in [3].
• examples/ZNE - notebooks from the paper [4], that show how temporally correlated dephasing noise impacts circuit folding techniques for zero noise extrapolation. Also has some examples of the recent filter function interface from the appendix of [4].

Citing mezze

If you use the SchWARMA tools in your research, please cite:

[1] Schultz, K., Quiroz, G., Titum, P., & Clader, B. D. (2020). SchWARMA: A model-based approach for time-correlated noise in quantum circuits. Physical Review Research 3, 033229. Available as an arXiv preprint arXiv:2010.04580.

If you use SchWARMA for noise injection, please cite:

[2] Murphy, A., Epstein, J., Quiroz, G., Schultz, K., Tewala, L., McElroy, K., ... & Sweeney, T. M. (2021). Universal Dephasing Noise Injection via Schrodinger Wave Autoregressive Moving Average Models. arXiv preprint arXiv:2102.03370.

If you experiment with the heavy-tailed (Levy-alpha stable) variants of SchWARMA, please cite:

[3] Clader, B. D., Trout, C. J., Barnes, J. P., Schultz, K., Quiroz, G., & Titum, P. (2021). Impact of correlations and heavy-tails on quantum error correction. Physical Review A 103, 052428. Available as an arXiv preprint arXiv:2101.11631.

The zero noise extrapolation examples are from

[4] Schultz, K., LaRose, R., Mari, A.,Quiroz, G., Shammah, N., Clader, B. D., and Zeng, W. J. (2022) Reducing the impact of time-correlated noise on zero-noise extrapolation. arXiv preprintarXiv:2201.11792

Tensorflow Quantum Branch

Note that most of the TensorFlow Quantum-based tools are now in the main branch, and tfq will only be for developing capabilities

To use the SchWARMA tools implemented using TensorFlow Quantum and Cirq:

• Install TensorFlow Quantum using the instructions here
• Fetch remote branches - git fetch
• Checkout the tfq branch of mezze - git checkout tfq
• Install mezze - pip install -e . (or similar)

This will resolve issues with dependencies that occur by using pip only