This repository contains the source code used in the computational experiments of the paper: Binary Control Pulse Optimization for Quantum Systems.
In our paper, we first apply GRAPE algorithm to solve the continuous relaxation of binary quantum control problems with SOS1 property. Then we apply two rounding techniques, sum-up rounding (SUR) and combinatorial integral approximation (CIA) to obtain the binary controls. Furthermore, we apply our improvement heuristic to improve the solutions.
If you use our code in your research, please cite our paper:
Binary Control Pulse Optimization for Quantum Systems
Xinyu Fei, Lucas T. Brady, Jeffrey Larson, Sven Leyffer, Siqian Shen
Quantum, 2023@article{fei2023binary, title={Binary control pulse optimization for quantum systems}, author={Fei, Xinyu and Brady, Lucas T and Larson, Jeffrey and Leyffer, Sven and Shen, Siqian}, journal={Quantum}, volume={7}, pages={892}, year={2023}, publisher={Verein zur F{\"o}rderung des Open Access Publizierens in den Quantenwissenschaften} }
There are three test instances in the paper:
- Energy minimization problem
- CNOT gate estimation problem
- Circuit compilation problem
For each instance, we solve the following optimization problems:
- Continuous relaxation
- Model with TV regularizer
- Rounding techniques
- Improvement heuristic
The instances are provided in the folder /example/.
- Python >= 3.8
- qiskit >= 0.29.0, scipy >= 1.6.2
- pycombina >= 0.3.2 (only for rounding)
- Developed version of Qutip (https://github.com/qutip/qutip).
To install pycombina, please refer to https://pycombina.readthedocs.io/en/latest/.
To install the developed version of Qutip, first clone the repository locally:
git clone https://github.com/xinyufei/Quantum-Control-qutip.gitThen, install Qutip by
python setup.py developAll the control results are stored in the folder example/control/. All the output control figures are stored in
example/figure/. The output files are stored in example/output/. One can change the
paths in files to change the positions.
Before starting your own experiments, we suggest deleting the above three folders to clear all the existing results.
First, change to the example file folder:
cd example/Continuous/To run an energy minimization problem with 4 qubits, randomly generated graph for Hamiltonian controllers, evolution time as 2, time steps as 40, constant initial control values 0.5, run:
python energy.py --n=4 --num_edges=2 --rgraph=1 --seed=1 \
--evo_time=2 --n_ts=40 --initial_type=CONSTANT --offset=0.5To run a CNOT estimation problem with evolution time 10, time steps 200, constant initial control values 0.5, run:
python CNOT.py --evo_time=10 --n_ts=200 --initial_type=CONSTANT --offset=0.5To run a circuit compilation problem on molecule LiH with evolution time 20, time steps 200, automatically generating target circuit, constant initial control values 0.5, and without SOS1 property, run
python Molecule.py --gen_target=1 --name=MoleculeVQE \
--molecule=LiH --qubit_num=4 --evo_time=20 --n_ts=100 \
--initial_type=CONSTANT --offset=0.5In circuit compilation problem, to penalize the SOS1 property, one can set the penalty parameter. For example, with the above setting adding penalty, run:
python Molecule.py --gen_target=1 --name=MoleculeVQE \
--molecule=LiH --qubit_num=4 --evo_time=20 --n_ts=100 \
--initial_type=CONSTANT --offset=0.5 --sum_penalty=0.1The output control files are stored in example/control/Continuous/. The output control figures are stored in
example/figure/Continuous/. The output files are stored in example/output/Continuous/.
First, change to the example file folder:
cd example/ADMM/To run ADMM for solving the model with TV regularizer on energy minimization problem, run:
python energy.py --n=4 --rgraph=1 --seed=1 --num_edges=2 --evo_time=2 --n_ts=40 --initial_type=WARM \
--initial_control="../control/Continuous/Energy4_evotime2.0_n_ts240_ptypeCONSTANT_offset0.5_instance1.csv" \
--alpha=1e-2 --rho=10 --max_iter_admm=50First, change to the example file folder:
cd example/Rounding/To run SUR on energy minimization problem, run:
python energy.py --n=4 --rgraph=1 --seed=1 --num_edges=2 --evo_time=2 --n_ts=40 \
--initial_control="../control/Continuous/Energy4_evotime2.0_n_ts40_ptypeCONSTANT_offset0.5_instance1.csv" \
--type=SURTo run CIA with min-up time constraints on energy minimization problem, run:
python energy.py --n=4 --rgraph=1 --seed=1 --num_edges=2 --evo_time=2 --n_ts=40 \
--initial_control="../control/Continuous/Energy4_evotime2.0_n_ts40_ptypeCONSTANT_offset0.5_instance1.csv" \
--type=minup --min_up=10First, change to the example file folder:
cd example/Trustregion/To run the improvement heuristic on energy minimization problem for improve the solutions with TV regularizer, run:
python energy.py --n=4 --rgraph=1 --seed=4 --evo_time=2 --n_ts=40 \
--initial_file="../control/Rounding/EnergyADMM4_evotime2.0_n_ts40_ptypeWARM_offset0.5_penalty0.01_ADMM_10.0_iter100_instance4_1_SUR.csv" \
--alpha=0.01 --tr_type="tv"To run the improvement heuristic on energy minimization problem for improve the solutions with min-up time constraints, run:
python energy.py --n=4 --num_edges=2 --rgraph=1 --seed=4 --evo_time=2 --n_ts=40 \
--initial_file="../control/Rounding/EnergyADMM4_evotime2.0_n_ts40_ptypeWARM_offset0.5_penalty0.01_ADMM_10.0_iter100_instance4_minup10_1.csv" \
--alpha=0.01 --tr_type="hard" --hard_type="minup" --min_up=10We thank Dr.Lucas Brady for providing the code used in the paper Optimal Protocols in Quantum Annealing and QAOA Problems (https://arxiv.org/pdf/2003.08952.pdf).
We refer to the paper Partial Compilation of Variational Algorithms for Noisy Intermediate-Scale Quantum Machines (https://arxiv.org/pdf/1909.07522.pdf) for generating the circuit compilation problem. Their code is presented at https://github.com/epiqc/PartialCompilation.
Xinyu Fei (xinyuf@umich.edu)
Xinyu Fei (xinyuf@umich.edu)
Siqian Shen (siqian@umich.edu)