RobChar - Robust Characterisation of Quantum Controls and Control Algorithms

Monte Carlo ROBustness CHARacterization and analysis of static controllers obtained using various gradient and gradient-free methods This is code accompanying the work "Statistically Characterising Robustness and Fidelity of Quantum Controls and Quantum Control Algorithms (2022)" here. The quantum model is an effective Heisenberg XXZ chain. In the paper, we don't enable the Z interaction available in the code.

To setup, please run

git clone git@qyber.black:spinnet/code-robchar.git
cd  code-robchar
pip3 install -r requirements.txt

Important/Likely Novel dependencies are sklearn, skquant and pytorch.

Top view Functionality

The code is mostly modular with every optimization algorithm being a subclass of LBFGS which is also an optimizer. The API to port your own optimizer is by overriding the run function. See qnewton.py for many examples. The noise_analysis.py file then lets you port your optimizer for collecting controllers for the spin chain

There is a separate class for handling cachable Monte Carlo simulations that only needs to be run once after the controllers have been collected.

Reproducing the figures

Data for reproducing the figures is provided with the code. Just running the python files individually or running the following

bash generate_all_figures.sh

will produce all the figures in the gray_scale_adjusted_paperfigs directory along with their grayscale representations in a gray sub-folder.

Running experiments to generate new controllers

To reproduce the data from the paper,

run

bash get_paper_data.sh

This will take a long time depending on your compute resources. But this should effectively populate/repopulate the experiments directory so that running the figure reproduction incantation above will produce figures based on your version of the experiments. On the other hand, if you look at get_paper_data.sh, you will realize that there are a number of distint sub-problems (e.g. a particular spin chain transition) delimited by a single line. Each is a separete experiment and can be run on its own.

Running experiments for other spin problems

Please have a look at get_paper_data.sh and parser.py. It should be straightforward to run a variant of the following with custom args to generate controllers for other problems not covered by the paper.

python noise_analysis.py --nspin `CHAIN_LENGTH` --outspin `OutputState` \
--inspin `InputState` \
--fid_threshold 0.1 --run_until_told_to_stop True \
--run_until_completion_its `num_objective_function_calls` --num_controllers 100

Locations

The code is developed and maintained on qyber\black at https://qyber.black/spinnet/code-robchar

This code is mirrored at

• https://github.com/qyber-black/code-robchar

The mirrors are only for convenience, accessibility and backup.

Citation

I. Khalid, C. A. Weidner, E. A. Jonckheere, S. G. Shermer, F. C. Langbein. RobChar: Robust Characterisation of Quantum Controls and Control Algorithms. Code, https://qyber.black/spinnet/code-robchar, 2022. [DOI:10.5281/zenodo.6891381]