Source code for openqaoa.backends.qaoa_vectorized

"""
Wavefunction simulator with methods focused on fast QAOA implementations.
Can easily be extended to do the full suite of operations in an ordinary simulator.
"""
from typing import Union, List, Tuple, Type, Optional
import numpy as np
from copy import copy
from scipy.sparse import csc_matrix, kron, diags
from scipy.sparse.linalg import expm

from .basebackend import QAOABaseBackendStatevector
from .gates_vectorized import VectorizedGateApplicator
from ..qaoa_components import QAOADescriptor, Hamiltonian
from ..qaoa_components.variational_parameters.variational_baseparams import (
    QAOAVariationalBaseParams,
)
from ..utilities import generate_uuid, round_value


# Pauli gates
constI = csc_matrix(np.eye(2))
constX = csc_matrix(np.array([[0, 1], [1, 0]]))
constY = csc_matrix(np.array([[0, -1j], [1j, 0]]))
constZ = csc_matrix(np.array([[1, 0], [0, -1]]))

# Single qubit gates
constH = csc_matrix((1 / np.sqrt(2)) * np.array([[1, 1], [1, -1]]))
constS = csc_matrix(np.array([[1, 0], [0, 1j]]))
constT = csc_matrix(np.array([[1, 0], [0, np.exp(1j * np.pi / 4)]]))

# Two-qubit gates
constCNOT = csc_matrix(
    np.array([[1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 0, 1], [0, 0, 1, 0]])
)
constCZ = csc_matrix(
    np.array([[1, 0, 0, 0], [0, 1, 0, 0], [0, 0, 1, 0], [0, 0, 0, -1]])
)

# Projection (measurement) operators
P0 = csc_matrix(np.array([[1, 0], [0, 0]]))
P1 = csc_matrix(np.array([[0, 0], [0, 1]]))


# Parametrised rotations
def RX(theta: float) -> csc_matrix:
    return csc_matrix(expm(-1j * theta * constX / 2))


def RY(theta: float) -> csc_matrix:
    return csc_matrix(expm(-1j * theta * constY / 2))


def RZ(theta: float) -> csc_matrix:
    return csc_matrix(expm(-1j * theta * constZ / 2))


def CR_Z(theta: float) -> csc_matrix:
    return kron(P0, constI, format="csc") + kron(P1, RZ(theta), format="csc")


def ZZ(theta: float) -> csc_matrix:
    return diags([1, np.exp(-1j * theta), np.exp(-1j * theta), 1], 0, format="csc")


def CPHASE(theta: float) -> csc_matrix:
    return diags([1, 1, 1, np.exp(1j * theta)], 0, format="csc")


def _get_perm(n_qubits: int, qubits: list) -> list:
    """
    Helper function for several methods below.

    Parameters
    ----------
    n_qubits:
        total number of qubits in the register

    qubits:
        the qubits being permuted to the beginning of the register

    Returns
    -------
    perm:
        the permutation to apply to the wavefunction to bring the
        active qubits to the start of the register

    perminv:
        the inverse permutation, taking them back to where they belong
        from the beginning
    """

    # Get modified indices of qubits to coincide with indexing from the right
    qubits = [np.arange(n_qubits)[::-1][i] for i in qubits]

    # Permute the active qubits to the beginning of the register
    other_qubits = list(set(np.arange(n_qubits)) - set(qubits))
    perm = qubits + other_qubits
    perminv = list(np.argsort(perm))

    return perm, perminv


def _permute_qubits(wavefn: np.ndarray, perm: list) -> np.ndarray:
    """
    Reorganises the wavefunction components according
    to the specified qubit permutation.

    Parameters
    ----------
    wavefn:
        The current wavefunction of the register

    perm:
        The permutation according to which the
        register is to be reorganised.

    Returns
    -------
    wavefn:
        The permuted wavefunction.
    """

    # Permute according to specified perm
    wavefn = np.transpose(wavefn, perm)

    return wavefn


def _build_cost_hamiltonian(
    n_qubits: int, cost_hamiltonian: Type[Hamiltonian]
) -> np.array:
    """
    Builds the cost Hamiltonian as a vector, since it is diagonal.
    Output is an ndarray of shape [2]*n_qubits,
    for use in the `expectation_value` method.

    Parameters
    ----------
    n_qubits:
        number of qubits

    cost_hamiltonian:
        Hamiltonian object containing information about
        single/2-qubit terms and their weights.

    Returns
    -------
    ham_op:
        the Hamiltonian as a diagonal matrix reshaped
        to a [2]*n_qubits dimensional array

    """

    bias_qubits = cost_hamiltonian.qubits_singles
    biases = cost_hamiltonian.single_qubit_coeffs
    pairs = cost_hamiltonian.qubits_pairs
    coeffs = cost_hamiltonian.pair_qubit_coeffs

    terms = cost_hamiltonian.terms
    weights = cost_hamiltonian.coeffs

    # Check for non-classical terms
    cost_ham_pauli_str_lst = [term.pauli_str for term in terms]
    for term in cost_ham_pauli_str_lst:
        if str(term) != "Z" and str(term) != "ZZ":
            raise Exception(
                f"Currently, only classical cost Hamiltonians"
                "that consists of 'Z' and 'ZZ' terms are supported,"
                "but a '{term}' term was encountered."
            )

    ## ZZ operator on first two qubits, identity on all others to the right
    # Diagonal matrix (vector) for cost function
    iden_plus = np.ones(2 ** (n_qubits - 2), dtype=int)
    iden_minus = -1 * np.ones(2 ** (n_qubits - 1), dtype=int)
    ZZ_op = np.hstack((iden_plus, np.hstack((iden_minus, iden_plus))))
    ZZ_op.shape = [2] * n_qubits

    ## Z operator on first qubit, identity on all others to the right
    Z_op = np.hstack(
        (
            np.ones(2 ** (n_qubits - 1), dtype=int),
            -1 * np.ones(2 ** (n_qubits - 1), dtype=int),
        )
    )
    Z_op.shape = [2] * n_qubits

    ham_op = np.zeros([2] * n_qubits)
    for j in range(len(pairs)):
        perm, perminv = _get_perm(n_qubits, pairs[j].qubit_indices)
        ZZ_op = _permute_qubits(ZZ_op, perminv)

        ham_op += coeffs[j] * ZZ_op

        ZZ_op = _permute_qubits(ZZ_op, perm)

    for j in range(len(bias_qubits)):
        perm, perminv = _get_perm(n_qubits, [bias_qubits[j].qubit_indices])
        Z_op = _permute_qubits(Z_op, perminv)

        ham_op += biases[j] * Z_op

        Z_op = _permute_qubits(Z_op, perm)

    # add the constant term from the hamiltonian
    ham_op += cost_hamiltonian.constant

    return ham_op


[docs]class QAOAvectorizedBackendSimulator(QAOABaseBackendStatevector): r""" A simulator class for quantum circuits, oriented to QAOA, and more generally unitaries generated by Hamiltonians which consists of sums of Pauli strings. Works by translating the actions of single and two-Pauli rotation gates into permutations of wavefunction coefficients, obtained by slicing the (2, 2, ..., 2)-shaped wavefunction. Procedure: * Decompose rotation matrices into sum of identity and Pauli matrices with Euler's formula. * Compute the action of the Pauli matrices * Pauli X matrix : Flip coefficients with 0 at i-th qubit with coefficients with 1 at i-th qubit * Pauli Y matrx : Multiply 1j to everything. Flip coefficients with 0 at i-th qubit with coefficients with 1 at i-th qubit, and multiply -1 to coefficients with 0 at i-th qubit. * Pauli Z matrix : multiply -1 to coefficients with 0 at i-th qubit. * Obtain final wavefunction by summing up :math:`\sin(\theta/2)* \textit{original wavefunction} - 1j*\cos(\theta/2)*\textit{processed wavefunction}`. Qubit labelling begins from the right, so that the right-most qubit has label 0, and the left-most has label n_qubits-1. Parameters ---------- qaoa_descriptor: QAOADescriptor An object of the class ``QAOADescriptor`` which contains information on circuit construction and depth of the circuit. prepend_state: np.array The initial state of the circuit (before Hadamards). An array of shape :math:`(2^{n_qubits},)` or (2, 2, ..., 2). Defaults to ``[1,0,...,0]`` if ``None``. append_state: np.array A unitary matrix of shape :math:`(2^{self.n_qubits}, 2^{self.n_qubits})`, to be multiplied to the output state. init_hadamard: bool Whether to apply Hadamard gates to the beginning of the QAOA part of the circuit. cvar_alpha: float Conditional Value-at-Risk (CVaR) - a measure that takes into account only the tail of the probability distribution arising from the circut's count dictionary. Must be between 0 and 1. Check https://arxiv.org/abs/1907.04769 for further details. """ def __init__( self, qaoa_descriptor: QAOADescriptor, prepend_state: Optional[Union[np.ndarray, List[complex]]], append_state: Optional[Union[np.ndarray, List[complex]]], init_hadamard: bool, cvar_alpha: float = 1, ): assert ( cvar_alpha == 1 ), "Please use the shot-based simulator for simulations with cvar_alpha < 1" QAOABaseBackendStatevector.__init__( self, qaoa_descriptor, prepend_state, append_state, init_hadamard, cvar_alpha, ) # Build the Hamiltonian operator as an array self.ham_op = _build_cost_hamiltonian(self.n_qubits, self.cost_hamiltonian) if self.n_qubits > 0: self.wavefn = np.zeros((2**self.n_qubits,), dtype=complex) self.wavefn[0] = 1 self.wavefn = self.wavefn.reshape([2] * self.n_qubits) else: self.wavefn = [] # Handle prepend state if self.prepend_state is not None: if isinstance(self.prepend_state, np.ndarray): if np.shape(self.prepend_state) == np.shape(self.wavefn): self.wavefn = self.prepend_state elif np.shape(self.prepend_state) == (2**self.n_qubits,): self.wavefn = self.prepend_state.reshape([2] * self.n_qubits) else: raise ValueError( "Error : Unsupported prepend_state specified." "Not of shape (2**n,) or (2, 2, ..., 2))." ) else: raise ValueError( "Error : Unsupported prepend_state specified. Not an ndarray." ) # Handle append state if self.append_state is not None: if isinstance(self.append_state, np.ndarray) and np.shape( self.append_state ) == (2**self.n_qubits, 2**self.n_qubits): # check unitarity of append_state matrix if not np.allclose( np.eye(2**self.n_qubits), self.append_state.dot(self.append_state.conj().T), ): raise ValueError("append_state is not a unitary matrix") else: raise ValueError( "Unsupported append_state specified (Not an ndarray," "or not of shape (2**n, 2**n)." ) # Handle init_hadamard if self.init_hadamard: for i in range(self.n_qubits): self.apply_hadamard(i) # store the initialisation part of wavefunction self.wavefn_init = copy(self.wavefn) # Apply gate methods
[docs] def apply_x(self, qubit_1: int): r""" Applies the X gate on ``qubit_1`` in a vectorized way. Parameters ---------- qubit_1: Qubit index to apply gate. Returns ------- None """ rotation_angle = np.pi C = np.cos(rotation_angle / 2) S = -1j * np.sin(rotation_angle / 2) wfn = (C * self.wavefn) + ( S * np.flip(self.wavefn, self.n_qubits - qubit_1 - 1) ) self.wavefn = wfn
[docs] def apply_rx(self, qubit_1: int, rotation_angle: float): r""" Applies the RX($\theta$ = ``rotation_angle``) gate on ``qubit_1`` in a vectorized way. **Definition of RX($\theta$):** .. math:: RX(\theta) = \exp\left(-i \frac{\theta}{2} X\right) = \begin{pmatrix} \cos{\frac{\theta}{2}} & -i\sin{\frac{\theta}{2}} \\ -i\sin{\frac{\theta}{2}} & \cos{\frac{\theta}{2}} \end{pmatrix} Parameters ---------- qubit_1: Qubit index to apply gate. rotation_angle: Angle to be rotated. Returns ------- None """ C = np.cos(rotation_angle / 2) S = -1j * np.sin(rotation_angle / 2) wfn = (C * self.wavefn) + ( S * np.flip(self.wavefn, self.n_qubits - qubit_1 - 1) ) self.wavefn = wfn
[docs] def apply_ry(self, qubit_1: int, rotation_angle: float): r""" Applies the RY($\theta$ = ``rotation_angle``) gate on ``qubit_1`` in a vectorized way. **Definition of RY($\theta$):** .. math:: RY(\theta) = \exp\left(-i \frac{\theta}{2} Y\right) = \begin{pmatrix} \cos{\frac{\theta}{2}} & -\sin{\frac{\theta}{2}} \\ \sin{\frac{\theta}{2}} & \cos{\frac{\theta}{2}} \end{pmatrix} Parameters ---------- qubit_1: Qubit index to apply gate. rotation_angle: Angle to be rotated. Returns ------- None """ wfn = copy(self.wavefn) # multiply slices with i/-i slc_0 = tuple( 0 if i == self.n_qubits - qubit_1 - 1 else slice(None) for i in range(self.n_qubits) ) slc_1 = tuple( 1 if i == self.n_qubits - qubit_1 - 1 else slice(None) for i in range(self.n_qubits) ) wfn[slc_0] *= -1j wfn[slc_1] *= 1j C = np.cos(rotation_angle / 2) S = 1j * np.sin(rotation_angle / 2) self.wavefn = (C * self.wavefn) + ( S * np.flip(wfn, self.n_qubits - qubit_1 - 1) )
[docs] def apply_rz(self, qubit_1: int, rotation_angle: float): r""" Applies the RZ($\theta$ = ``rotation_angle``) gate on ``qubit_1`` in a vectorized way. **Definition of RZ($\theta$):** .. math:: RZ(\theta) = \exp\left(-i \frac{\theta}{2} Z\right) = \begin{pmatrix} e^{-i\frac{\theta}{2}} & 0 \\ 0 & e^{i \frac{\theta}{2}} \end{pmatrix} Parameters ---------- qubit_1: Qubit index to apply gate. rotation_angle: Angle to be rotated. Returns ------- None """ slc_0 = tuple( 0 if i == self.n_qubits - qubit_1 - 1 else slice(None) for i in range(self.n_qubits) ) slc_1 = tuple( 1 if i == self.n_qubits - qubit_1 - 1 else slice(None) for i in range(self.n_qubits) ) self.wavefn[slc_0] *= np.exp(-1j * rotation_angle / 2) self.wavefn[slc_1] *= np.exp(1j * rotation_angle / 2)
[docs] def apply_rxx(self, qubit_1: int, qubit_2: int, rotation_angle: float): r""" Applies the RXX($\theta$ = ``rotation_angle``) gate on ``qubit_1`` and ``qubit_2`` in a vectorized way. **Definition of RXX($\theta$):** .. math:: R_{XX}(\theta) = \exp\left(-i \frac{\theta}{2} X{\otimes}X\right) = \begin{pmatrix} \cos\left(\frac{\theta}{2}\right) & 0 & 0 & -i\sin\left(\frac{\theta}{2}\right) \\ 0 & \cos\left(\frac{\theta}{2}\right) & -i\sin\left(\frac{\theta}{2}\right) & 0 \\ 0 & -i\sin\left(\frac{\theta}{2}\right) & \cos\left(\frac{\theta}{2}\right) & 0 \\ -i\sin\left(\frac{\theta}{2}\right) & 0 & 0 & \cos\left(\frac{\theta}{2}\right) \end{pmatrix} Parameters ---------- qubit_1: First qubit index to apply gate. qubit_2: Second qubit index to apply gate. rotation_angle: Angle to be rotated. Returns ------- None """ # SH TODO : investigate if slicing 01 and 10 coefficients and swapping once them is faster than flipping twice C = np.cos(rotation_angle / 2) S = -1j * np.sin(rotation_angle / 2) wfn = (C * self.wavefn) + ( S * np.flip( np.flip(self.wavefn, self.n_qubits - qubit_1 - 1), self.n_qubits - qubit_2 - 1, ) ) self.wavefn = wfn
[docs] def apply_ryy(self, qubit_1: int, qubit_2: int, rotation_angle: float): r""" Applies the RYY($\theta$ = ``rotation_angle``) gate on ``qubit_1`` and ``qubit_2`` in a vectorized way. **Definition of RYY($\theta$):** .. math:: R_{YY}(\theta) = \exp\left(-i \frac{\theta}{2} X{\otimes}X\right) = \begin{pmatrix} \cos\left(\frac{\theta}{2}\right) & 0 & 0 & -i\sin\left(\frac{\theta}{2}\right) \\ 0 & \cos\left(\frac{\theta}{2}\right) & -i\sin\left(\frac{\theta}{2}\right) & 0 \\ 0 & -i\sin\left(\frac{\theta}{2}\right) & \cos\left(\frac{\theta}{2}\right) & 0 \\ -i\sin\left(\frac{\theta}{2}\right) & 0 & 0 & \cos\left(\frac{\theta}{2}\right) \end{pmatrix} Parameters ---------- qubit_1: First qubit index to apply gate. qubit_2: Second qubit index to apply gate. rotation_angle: Angle to be rotated. Returns ------- None """ wfn = copy(self.wavefn) slc_q1_0 = tuple( 0 if i == self.n_qubits - qubit_1 - 1 else slice(None) for i in range(self.n_qubits) ) slc_q1_1 = tuple( 1 if i == self.n_qubits - qubit_1 - 1 else slice(None) for i in range(self.n_qubits) ) slc_q2_0 = tuple( 0 if i == self.n_qubits - qubit_2 - 1 else slice(None) for i in range(self.n_qubits) ) slc_q2_1 = tuple( 1 if i == self.n_qubits - qubit_2 - 1 else slice(None) for i in range(self.n_qubits) ) wfn[slc_q1_0] *= 1j wfn[slc_q1_1] *= -1j wfn[slc_q2_0] *= -1j wfn[slc_q2_1] *= 1j C = np.cos(rotation_angle / 2) S = 1j * np.sin(rotation_angle / 2) self.wavefn = (C * self.wavefn) + ( S * np.flip( np.flip(wfn, self.n_qubits - qubit_1 - 1), self.n_qubits - qubit_2 - 1 ) )
[docs] def apply_rzz(self, qubit_1: int, qubit_2: int, rotation_angle: float): r""" Applies the RZZ($\theta$ = ``rotation_angle``) gate on ``qubit_1`` and ``qubit_2`` in a vectorized way. **Definition of RZZ($\theta$):** .. math:: RZZ(\theta) = \exp\left(-i \frac{\theta}{2} Z{\otimes}Z\right) = \begin{pmatrix} e^{-i \frac{\theta}{2}} & 0 & 0 & 0 \\ 0 & e^{i \frac{\theta}{2}} & 0 & 0 \\ 0 & 0 & e^{i \frac{\theta}{2}} & 0 \\ 0 & 0 & 0 & e^{-i \frac{\theta}{2}} \end{pmatrix} Parameters ---------- qubit_1: First qubit index to apply gate. qubit_2: Second qubit index to apply gate. rotation_angle: Angle to be rotated. Returns ------- None """ """ # Note : one can also slice the 01 and 10 elements: slc_pair01 = tuple(1 if i == self.n_qubits - qubit_1 - 1 else 0 if i == self.n_qubits - qubit_2 - 1 else slice(None) for i in range(self.n_qubits)) slc_pair10 = tuple(1 if i == self.n_qubits - qubit_2 - 1 else 0 if i == self.n_qubits - qubit_1 - 1 else slice(None) for i in range(self.n_qubits)) """ slc_pair00 = tuple( 0 if i in [ self.n_qubits - qubit_1 - 1, self.n_qubits - qubit_2 - 1, ] else slice(None) for i in range(self.n_qubits) ) slc_pair11 = tuple( 1 if i in [ self.n_qubits - qubit_1 - 1, self.n_qubits - qubit_2 - 1, ] else slice(None) for i in range(self.n_qubits) ) self.wavefn[slc_pair00] *= np.exp(-1j * rotation_angle) self.wavefn[slc_pair11] *= np.exp(-1j * rotation_angle) self.wavefn *= np.exp(1j * rotation_angle / 2)
[docs] def apply_rxy(self, qubit_1: int, qubit_2: int, rotation_angle: float): """ Applies the RXY($\theta$ = ``rotation_angle``) gate on ``qubit_1`` and ``qubit_2`` in a vectorized way. Parameters ---------- qubit_1: First qubit index to apply gate. qubit_2: Second qubit index to apply gate. rotation_angle: Angle to be rotated. Returns ------- None """ wfn = copy(self.wavefn) # Action of Y part slc_q2_0 = tuple( 0 if i == self.n_qubits - qubit_2 - 1 else slice(None) for i in range(self.n_qubits) ) slc_q2_1 = tuple( 1 if i == self.n_qubits - qubit_2 - 1 else slice(None) for i in range(self.n_qubits) ) wfn[slc_q2_0] *= -1j wfn[slc_q2_1] *= 1j C = np.cos(rotation_angle / 2) S = 1j * np.sin(rotation_angle / 2) self.wavefn = (C * self.wavefn) + ( S * np.flip( np.flip(wfn, self.n_qubits - qubit_1 - 1), self.n_qubits - qubit_2 - 1 ) )
[docs] def apply_rzx(self, qubit_1: int, qubit_2: int, rotation_angle: float): """ Applies the RZX($\theta$ = ``rotation_angle``) gate on ``qubit_1`` and ``qubit_2`` in a vectorized way. Parameters ---------- qubit_1: First qubit index to apply gate. qubit_2: Second qubit index to apply gate. rotation_angle: Angle to be rotated. Returns ------- None """ wfn = copy(self.wavefn) # Action of Z part slc_q2_0 = tuple( 0 if i == self.n_qubits - qubit_1 - 1 else slice(None) for i in range(self.n_qubits) ) slc_q2_1 = tuple( 1 if i == self.n_qubits - qubit_1 - 1 else slice(None) for i in range(self.n_qubits) ) wfn[slc_q2_0] *= 1 wfn[slc_q2_1] *= -1 C = np.cos(rotation_angle / 2) S = -1j * np.sin(rotation_angle / 2) self.wavefn = (C * self.wavefn) + ( S * np.flip(wfn, self.n_qubits - qubit_2 - 1) )
[docs] def apply_ryz(self, qubit_1: int, qubit_2: int, rotation_angle: float): """ Applies the RYZ($\theta$ = ``rotation_angle``) gate on ``qubit_1`` and ``qubit_2`` in a vectorized way. Parameters ---------- qubit_1: First qubit index to apply gate. qubit_2: Second qubit index to apply gate. rotation_angle: Angle to be rotated. Returns ------- None """ wfn = copy(self.wavefn) # Action of Y part slc_q1_0 = tuple( 0 if i == self.n_qubits - qubit_1 - 1 else slice(None) for i in range(self.n_qubits) ) slc_q1_1 = tuple( 1 if i == self.n_qubits - qubit_1 - 1 else slice(None) for i in range(self.n_qubits) ) wfn[slc_q1_0] *= -1j wfn[slc_q1_1] *= 1j # Action of Z part slc_q2_0 = tuple( 0 if i == self.n_qubits - qubit_2 - 1 else slice(None) for i in range(self.n_qubits) ) slc_q2_1 = tuple( 1 if i == self.n_qubits - qubit_2 - 1 else slice(None) for i in range(self.n_qubits) ) wfn[slc_q2_0] *= 1 wfn[slc_q2_1] *= -1 C = np.cos(rotation_angle / 2) S = 1j * np.sin(rotation_angle / 2) self.wavefn = (C * self.wavefn) + ( S * np.flip(wfn, self.n_qubits - qubit_1 - 1) )
[docs] def apply_hadamard(self, qubit_1: int): """ Applies the Hadamard gate on ``qubit_1`` in a vectorized way. Only used when ``init_hadamard`` is true. Parameters ---------- qubit_1: First qubit index to apply gate. Returns ------- None """ # TODO : Combine init_hadamard and prepend_state into one. # vectorized hadamard gate, for when init_hadamard = True wfn = copy(self.wavefn) slc_0 = tuple( 0 if i == self.n_qubits - qubit_1 - 1 else slice(None) for i in range(self.n_qubits) ) slc_1 = tuple( 1 if i == self.n_qubits - qubit_1 - 1 else slice(None) for i in range(self.n_qubits) ) wfn[slc_1] *= -1 wfn[slc_0] += self.wavefn[slc_1] wfn[slc_1] += self.wavefn[slc_0] self.wavefn = wfn / np.sqrt(2)
[docs] def qaoa_circuit(self, params: Type[QAOAVariationalBaseParams]): """ Executes the entire QAOA circuit, with angles specified within ``params``. Steps: 1) Creates a (2,...,2) dimensional matrix that represents a 2**n dimensional wavefunction 2) Modify it according to the ``prepend_state`` option. 3) Modify it according to ``init_hadamard`` option. 4) Modify it according to list of gates in ``params``. 5) Modify it accoding to ``append_state`` option. Parameters ---------- params: ``QAOAVariationalBaseParams`` object that contains rotation angles and gates to be applied. Returns ------- None """ gates_applicator = VectorizedGateApplicator() # generate a job id for the wavefunction evaluation self.job_id = generate_uuid() # reset the wavefunction back to its initialisation state self.reset_circuit() # Assign angles and apply gates self.assign_angles(params) low_level_gate_list = [] for each_gate in self.abstract_circuit: low_level_gate_list.extend(each_gate.decomposition("trivial")) for each_tuple in low_level_gate_list: gate = each_tuple[0](gates_applicator, *each_tuple[1]) gate.apply_gate(self) # Handle append state if self.append_state is not None: # Flatten (2,...,2) shaped wfn into a 2**n-dim column vector before multiplying with unitary matrix, ... self.wavefn = np.matmul(self.append_state, self.wavefn.flatten()) # then re-shape it back to (2,...,2) self.wavefn = self.wavefn.reshape([2] * self.n_qubits)
[docs] def wavefunction(self, params: Type[QAOAVariationalBaseParams] = None) -> list: """ Get the wavefunction of the state produced by the parametric circuit. Parameters ---------- params: The QAOA parameters - an object of one of the parameter classes, containing hyperparameters and variable parameters. Returns ------- wf: A list of the wavefunction amplitudes. """ self.qaoa_circuit(params) self.wavefn.shape = 2**self.n_qubits self.measurement_outcomes = self.wavefn.flatten() # Make format same as ProjectQ wf = [(component) for component in self.wavefn] return wf
@round_value def expectation(self, params: Type[QAOAVariationalBaseParams]) -> float: """ Call the execute function on the circuit to compute the expectation value of the Quantum Circuit w.r.t cost operator Returns ------- exp_val: The expectation value of the cost function wrt the state generated by the circuit. """ self.qaoa_circuit(params) # Reshape wavefunction wavefn_ = self.wavefn self.measurement_outcomes = wavefn_.flatten() # Compute the expectation value and its standard deviation ham_wf = self.ham_op * wavefn_ exp_val = np.real(np.vdot(wavefn_, ham_wf)) out = exp_val return out @round_value def expectation_w_uncertainty( self, params: Type[QAOAVariationalBaseParams] ) -> Tuple[float, float]: """ Call the execute function on the circuit to compute the expectation value of the ``QuantumCircuit`` w.r.t cost operator along with its uncertainty Returns ------- exp_val: The expectation value of the cost function wrt the state generated by the circuit. std_dev: The standard deviation of the cost function wrt the state generated by the circuit. """ self.qaoa_circuit(params) # Reshape wavefunction wavefn_ = self.wavefn self.measurement_outcomes = wavefn_.flatten() # Compute the expectation value and its standard deviation ham_wf = self.ham_op * wavefn_ exp_val = np.real(np.vdot(wavefn_, ham_wf)) exp_val_sq = np.real(np.vdot(ham_wf, ham_wf)) std_dev = (exp_val_sq - exp_val**2) ** 0.5 out = exp_val, std_dev return out
[docs] def reset_circuit(self): """ Reset the circuit by resetting the wavefunction. """ self.wavefn = copy(self.wavefn_init)
[docs] def circuit_to_qasm(self): """ A method to convert the entire QAOA ``QuantumCircuit`` object into a OpenQASM string """ raise NotImplementedError()