This repository contains the Python implementation of a flexible and efficient gate set tomography technique based on low-rank approximations, as described in the following research paper:
Title: Compressive gate set tomography
Authors: Raphael Brieger, Ingo Roth, and Martin Kliesch
Links: https://arxiv.org/abs/2112.05176, https://journals.aps.org/prxquantum/pdf/10.1103/PRXQuantum.4.010325
Flexible characterization techniques that identify and quantify experimental imperfections under realistic assumptions are crucial for the development of quantum computers. Gate set tomography is a characterization approach that simultaneously and self-consistently extracts a tomographic description of the implementation of an entire set of quantum gates, as well as the initial state and measurement, from experimental data. Obtaining such a detailed picture of the experimental implementation is associated with high requirements on the number of sequences and their design, making gate set tomography a challenging task even for only two qubits. In this work, we show that low-rank approximations of gate sets can be obtained from significantly fewer gate sequences and that it is sufficient to draw them randomly. Such tomographic information is needed for the crucial task of dealing with coherent noise. To this end, we formulate the data processing problem of gate set tomography as a rank-constrained tensor completion problem. We provide an algorithm to solve this problem while respecting the usual positivity and normalization constraints of quantum mechanics by using second-order geometrical optimization methods on the complex Stiefel manifold. Besides the reduction in sequences, we demonstrate numerically that the algorithm does not rely on structured gate sets or an elaborate circuit design to robustly perform gate set tomography. Therefore, it is more flexible than traditional approaches. We also demonstrate how coherent errors in shadow estimation protocols can be mitigated using estimates from gate set tomography.
This Python package allows users to perform gate set tomography on quantum devices using low-rank approximations, which reduces the number of gate sequences and the post processing time required. The standard sequence design needed for compressive GST is given by random sequences of short depth, and the python package provides functions that generate those sequences based on high level parameters such as the number of gates and the system size. In its standard configuration, the mGST algorithm assumes no prior knowledge about the gate set parameters and uses local optimization coupled with random initializations until a high quality fit is obtained. Alternatively, an initial guess for the initialization can be supplied, which can lead to large reductions in runtime.
For single qubit gate set tomography we recommend using 100 sequences with at least 1000 shots each. For two qubits with model rank up to 4, we recommend 400 sequences and also at least 1000 shots each. The runtime varies for different gate sets and ranks, but is expected to be on the order of a minute for a single qubit and on the order of an hour for two qubits. Details can be found in the linked publication.
The folder tutorials/sample_project contains the blueprint for automatically running mGST from the command line.
In order to use make use of this functionality, the following steps should be performed: 1) Copy the entire "sample_project" folder to a desired location and rename to identify the current experiment. 2) Replace the file "synthetic_data.txt" in the "exp_data" subfolder with a text file containting experimental data. 3) Edit the file "exp_description" in the "exp_design" subfolder to define the parameters in your experiment. 4) Replace the file "sequences.csv" with the gate sequences used in your current experiment. 5) Open a terminal in the main project folder and run the command "python3 run_mGST.py".
The mGST-Algorithm should now run and give updates on the current state of the optimization. After an estimate is found, "y/n" promts will guide through the next steps considering whether bootstrapping error bars and tex-reports should be generated. Gate error tables as well as plots of gate parametrizations are saved in the subfolder "mGST_results", and the full late-report is saved in the "report"-subfolder.
To install the package, follow these steps:
git clone git@github.com:rabrie/mGST.gitcd mGSTpip install .Now you can import the package in your Python scripts and use it. For examples and usage instructions, please refer to the provided Jupyter notebook.