Code used to produce the results of the paper G. García-Pérez, M. A. C. Rossi, B. Sokolov, F. Tacchino, P. K. Barkoutsos, G. Mazzola, I. Tavernelli, S. Maniscalco, "Learning to Measure: Adaptive Informationally Complete Generalized Measurements for Quantum Algorithms", PRX Quantum 2, 040342 (2021).
The code uses qiskit 0.23 for simulating a near term quantum device, generating qubit Hamiltonian for small molecules and running VQE.
NOTE: The simulations are quite demanding, especially for large molecules. The data was generated using a HPC cluster, and is openly available on Zenodo.org.
This repository requires Python 3 (tested with Python 3.8.5). Since we need to patch the qiskit aqua module in order to add the JKMN mapping, we recommend creating a virtual environment.
Clone the repository and move to the folder
git clone https://github.com/matteoacrossi/adapt_ic-povm.git
cd adapt_ic-povm
Install the prerequisites with
pip install -r requirements.txt
Apply the patch to add the JKMN mapping
pypatch apply neven.patch qiskit
The data used to reproduce the results presented in the paper is available at https://doi.org/10.5281/zenodo.5759030. Download the
file data.tar and extract it in the repository folder:
wget https://zenodo.org/record/5759030/files/data.tar
tar -xvf data.tar
A folder data/ will be created.
The example Hamiltonians are generated with python preprare_chemistry_hamiltonians.py. It will take a considerable amount of time and a workstation or cluster is recommended.
The molecules are defined in the file preprare_chemistry_hamiltonians.py.
The file data/hamiltonians.pickle contains a list of example chemistry Hamiltonians for the H2, LiH and H2O molecules,
with different bases, symmetries and fermion-to-qubit mappings. For each Hamiltonian, the exact ground state and an approximated
ground state found with VQE (using qiskit 0.23.6 default settings) are stored in the file.
Each entry in the file is a dictionary of the form:
{'molecule': 'H2',
'basis': 'sto3g',
'two_qubit_reduction': False,
'z2symmetry_reduction': False,
'freeze_core': False,
'mapping': 'parity',
'operator': <qiskit.aqua.operators.legacy.weighted_pauli_operator.WeightedPauliOperator at 0x2b3a9078b9a0>,
'qubits': 4,
'num_paulis': 15,
'vqe_circuit': <qiskit.circuit.library.n_local.real_amplitudes.RealAmplitudes at 0x2b3a8febd970>,
'vqe_value': -1.842665192696192,
'exact_circuit': <qiskit.circuit.quantumcircuit.QuantumCircuit at 0x2b3a8f0747f0>,
'exact_value': -1.842686681905733,
'name': '4q H2 parity'}Similarly, the file data/hamiltonians_h_chain.pickle contains a list of Hamiltonians for hydrogen chains of various length. The file can be generated with python prepare_h_chain_hamiltonians.py.
The list of hamiltonians is shown in Hamiltonians_summary.ipynb.
The simulations can be executed with run_simulation.py. The file takes a number of arguments as input
usage: run_simulation.py [-h] [--file FILE] [--hamiltonian HAMILTONIAN] [--samples SAMPLES] [--shots SHOTS [SHOTS ...]] [--method METHOD] [--outfile OUTFILE] [--exact] [--counts]
optional arguments:
-h, --help show this help message and exit
--file FILE The file with the Hamiltonians (hamiltonians.pickle)
--hamiltonian HAMILTONIAN The id of the Hamiltonian
--samples SAMPLES How many samples to take
--shots SHOTS [SHOTS ...] The shots to use for the measurement (multiple amounts of shots can be passed)
--method METHOD One of the methods: SIC-POVM, Pauli, Grouped_Pauli, Grad-POVM, Google-POVM, Grad-Google-POVM
--outfile OUTFILE File where to save the results
--exact Use the exact diagonalization state
--counts Store the measurement counts to file (for tomography)
The output files are in the JSON lines format:
{"qubits": 4, "true": -1.8426866160309918, "estimate": -1.8394723164727411, "estimated_error": 0.01613556904575243, "error": 0.0032142995582506995, "circuits": 15, "shots_per_circuit": 66, "shots": 990, "time_qc": 2.879150152206421, "time_post": 0.5117971897125244, "method": "Pauli", "counts": null, "name": "4q H2 jordan_wigner", "commit": "d59fe99", "exact_ground_state": false, "id": "MzZya4nhC6fQdvZsob332H", "timestamp": 1625040177.894622}
{"qubits": 4, "true": -1.8426866160309918, "estimate": -1.8418662180482184, "estimated_error": 0.01523071106596722, "error": 0.0008203979827734464, "circuits": 15, "shots_per_circuit": 66, "shots": 990, "time_qc": 2.938889980316162, "time_post": 0.522057294845581, "method": "Pauli", "counts": null, "name": "4q H2 jordan_wigner", "commit": "d59fe99", "exact_ground_state": false, "id": "Xd9n4GGivADMxi8fT6CuZJ", "timestamp": 1625040177.956557}
...and can be easily processed with Pandas.
A convenience script process_raw_data.py joins multiple files, adds Hamiltonian information and saves the final dataset to a binary .feather file. For example:
python process_raw_data.py 'raw_data/*.txt' --hamiltonians hamiltonians.pickle -o data/chemistry_data.feather
For producing the results related to k-RDM tomography, all the counts coming from the measurements need to be stored. This can be achieved with the --counts flag of run_simulation.py.
The data used for the paper is stored in the JSON Lines file counts_data.txt. The data can be processed using python tomography_script.py.
The notebook Figures.ipynb generates all the figures contained in the published paper from the data available at https://doi.org/10.5281/zenodo.5759030.
G. García-Pérez, M. A. C. Rossi, B. Sokolov, F. Tacchino, P. K. Barkoutsos, G. Mazzola, I. Tavernelli, S. Maniscalco, "Learning to Measure: Adaptive Informationally Complete Generalized Measurements for Quantum Algorithms", PRX Quantum 2, 040342 (2021).
@misc{garciaperez2021learning,
title = {Learning to Measure: Adaptive Informationally Complete Generalized Measurements for Quantum Algorithms},
author = {Garc\'{\i}a-P\'erez, Guillermo and Rossi, Matteo A.C. and Sokolov, Boris and Tacchino, Francesco and Barkoutsos, Panagiotis Kl. and Mazzola, Guglielmo and Tavernelli, Ivano and Maniscalco, Sabrina},
journal = {PRX Quantum},
volume = {2},
issue = {4},
pages = {040342},
numpages = {17},
year = {2021},
month = {Nov},
publisher = {American Physical Society},
doi = {10.1103/PRXQuantum.2.040342},
url = {https://link.aps.org/doi/10.1103/PRXQuantum.2.040342}
}