qiu-mps-initializer

qiu-mps-initializer prepares quantum states approximately with layers of one- and two-qubit gates obtained from matrix product states (MPS), following Ran, "Encoding of matrix product states into quantum circuits of one- and two-qubit gates", Phys. Rev. A 101, 032310 (2020). Each layer acts on neighboring qubits only and costs at most 3 (n - 1) CNOT gates on n qubits, so a few layers give circuits much shallower than exact state preparation, whose CNOT count grows as 2**n, at the cost of an approximation. Every preparation reports its error.

Within the monorepo, it builds on qiu-qiskit-encore, which validates its states, and on qiu-quantum-computing, which prepares single-qubit states exactly and provides the states of intensity signals of qiu-signals; it complements the exact state preparation there. Version 0.3 carries the functionality of version 0.2 over onto these packages, see the User Guide.

Installation

pip install qiu-mps-initializer

Inside the monorepo, it is installed with all other packages by uv sync --all-packages.

Quick start

import numpy as np
from qiskit.quantum_info import Statevector
from qiu_mps_initializer.state_preparation import mps_state_preparation

rng = np.random.default_rng(0)
state = rng.normal(size=32) + 1j * rng.normal(size=32)
state /= np.linalg.norm(state)

preparation = mps_state_preparation(state, max_layers=8)
prepared = Statevector(preparation.circuit).data
assert preparation.num_layers == 8
assert np.isclose(preparation.error, np.linalg.norm(prepared - state))
assert abs(np.vdot(state, prepared)) ** 2 > 0.99  # the fidelity

Where next

• The User Guide explains the MPS layers, the multi-layer preparation, its error and its convergence.
• The Examples show worked use cases, from the fidelity of increasing numbers of layers to preparing the state of an intensity signal.
• The API Reference documents every module, class and function.