A research tool to explore Quantum Error Correction (QEC), primarily surface codes.
[Error] we're working on the documentation of this project, please wait for a formal release (1.0.0) before you want to use this project.
See the QEC-Playground Documentation: Installation for the detailed instructions. A brief example is below.
# Download the Blossom V Library [Optional]
wget -c https://pub.ist.ac.at/~vnk/software/blossom5-v2.05.src.tar.gz -O - | tar -xz
cp -r blossom5-v2.05.src/* backend/blossomV/
rm -r blossom5-v2.05.src
# Install the Python Dependencies [Optional]
sudo apt install python3 python3-pip
pip3 install networkx
# Install the Rust Toolchain
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
source ~/.bashrc # this will add `~/.cargo/bin` to path
cd backend/rust/
cargo build --release
cd ../../
See the QEC-Playground Documentation: CLI for the detailed instructions. A brief example use case is below.
Run cargo run --release -- --help
under backend/rust/
folder to get all provided commands of backend program.
The option --help
prints out the information of this command, which can be helpful to find subcommands as well as to understand the purpose of each option.
An example output is below.
QECPlayground 0.1.6
Yue Wu <[email protected]>, Namitha Liyanage ([email protected])
Quantum Error Correction Playground
USAGE:
qecp <SUBCOMMAND>
OPTIONS:
-h, --help Print help information
-V, --version Print version information
SUBCOMMANDS:
fpga_generator fpga_generator
help Print this message or the help of the given subcommand(s)
server HTTP server for decoding information
test testing features
tool tools
To run a simulation to benchmark the logical error rate of decoder, run cargo run --release -- tool benchmark --help
. An example output is below.
qecp-tool-benchmark 0.1.6
benchmark surface code decoders
USAGE:
qecp tool benchmark [OPTIONS] <dis> <nms> <ps>
ARGS:
<dis> [di1,di2,di3,...,din] code distance of vertical axis
<nms> [nm1,nm2,nm3,...,nmn] number of noisy measurement rounds, must have exactly the
same length as `dis`; note that a perfect measurement is always capped at the end,
so to simulate a single round of perfect measurement you should set this to 0
<ps> [p1,p2,p3,...,pm] p = px + py + pz unless noise model has special interpretation of
this value
OPTIONS:
--bias_eta <bias_eta>
bias_eta = pz / (px + py) and px = py, px + py + pz = p. default to 1/2, which means px
= pz = py [default: 0.5]
......
For example, to test code-distance-3 standard CSS surface code with depolarizing physical error rates 3%, 2% and 1% only on data qubits (i.e. perfect stabilizer measurements) using the default decoder (MWPM decoder), run:
cargo run --release -- tool benchmark [3] [0] [3e-2,2e-2,1e-2]
An example result is below.
format: <p> <di> <nm> <total_repeats> <qec_failed> <error_rate> <dj> <confidence_interval_95_percent> <pe>
0.03 3 0 567712 10000 0.01761456513161603 3 1.9e-2 0
0.02 3 0 1255440 10000 0.007965334862677627 3 2.0e-2 0
0.01 3 0 4705331 10000 0.002125248999485902 3 2.0e-2 0
See CHANGELOG.md
Yue Wu ([email protected]): implement 3D GUI. design and implement interactive tutorial. propose and implement naïve decoder. implement MWPM decoder. Implement different variants of surface code and different decoders (see change log 2020.11.8 - 2022.3.20). The major developer and maintainer of this repository.
Guojun Chen: collaborator of CPSC 559 course project: design GUI. design and implement machine learning based weight optimized MWPM decoder.
Namitha Godawatte Liyanage: implement approximate MWPM decoder and FPGA related functionalities.
Neil He: bind library to Python.
When using QEC-Playground for research, please cite:
TODO: arXiv link for related papers (probably the fusion blossom paper)