PauliTwoDesign

class qiskit.circuit.library.PauliTwoDesign(num_qubits=None, reps=3, seed=None, insert_barriers=False, name='PauliTwoDesign')

Bases: TwoLocal

The Pauli Two-Design ansatz.

This class implements a particular form of a 2-design circuit [1], which is frequently studied in quantum machine learning literature, such as e.g. the investigating of Barren plateaus in variational algorithms [2].

The circuit consists of alternating rotation and entanglement layers with an initial layer of $\sqrt{H} = RY(\pi/4)$ gates. The rotation layers contain single qubit Pauli rotations, where the axis is chosen uniformly at random to be X, Y or Z. The entanglement layers is compromised of pairwise CZ gates with a total depth of 2.

For instance, the circuit could look like this (but note that choosing a different seed yields different Pauli rotations).

     ┌─────────┐┌──────────┐       ░ ┌──────────┐       ░  ┌──────────┐
q_0: ┤ RY(π/4) ├┤ RZ(θ[0]) ├─■─────░─┤ RY(θ[4]) ├─■─────░──┤ RZ(θ[8]) ├
     ├─────────┤├──────────┤ │     ░ ├──────────┤ │     ░  ├──────────┤
q_1: ┤ RY(π/4) ├┤ RZ(θ[1]) ├─■──■──░─┤ RY(θ[5]) ├─■──■──░──┤ RX(θ[9]) ├
     ├─────────┤├──────────┤    │  ░ ├──────────┤    │  ░ ┌┴──────────┤
q_2: ┤ RY(π/4) ├┤ RX(θ[2]) ├─■──■──░─┤ RY(θ[6]) ├─■──■──░─┤ RX(θ[10]) ├
     ├─────────┤├──────────┤ │     ░ ├──────────┤ │     ░ ├───────────┤
q_3: ┤ RY(π/4) ├┤ RZ(θ[3]) ├─■─────░─┤ RX(θ[7]) ├─■─────░─┤ RY(θ[11]) ├
     └─────────┘└──────────┘       ░ └──────────┘       ░ └───────────┘

Examples

from qiskit.circuit.library import PauliTwoDesign
circuit = PauliTwoDesign(4, reps=2, seed=5, insert_barriers=True)
circuit.draw('mpl')

References

[1]: Nakata et al., Unitary 2-designs from random X- and Z-diagonal unitaries.

arXiv:1502.07514

[2]: McClean et al., Barren plateaus in quantum neural network training landscapes.

arXiv:1803.11173

Parameters

• num_qubits (int | None) – The number of qubits of the two-local circuit.
• rotation_blocks – The gates used in the rotation layer. Can be specified via the name of a gate (e.g. 'ry') or the gate type itself (e.g. RYGate). If only one gate is provided, the gate same gate is applied to each qubit. If a list of gates is provided, all gates are applied to each qubit in the provided order. See the Examples section for more detail.
• entanglement_blocks – The gates used in the entanglement layer. Can be specified in the same format as rotation_blocks.
• entanglement – Specifies the entanglement structure. Can be a string ('full', 'linear', 'reverse_linear', 'circular' or 'sca'), a list of integer-pairs specifying the indices of qubits entangled with one another, or a callable returning such a list provided with the index of the entanglement layer. Default to 'full' entanglement. Note that if entanglement_blocks = 'cx', then 'full' entanglement provides the same unitary as 'reverse_linear' but the latter option has fewer entangling gates. See the Examples section for more detail.
• reps (int) – Specifies how often a block consisting of a rotation layer and entanglement layer is repeated.
• skip_unentangled_qubits – If True, the single qubit gates are only applied to qubits that are entangled with another qubit. If False, the single qubit gates are applied to each qubit in the ansatz. Defaults to False.
• skip_final_rotation_layer – If False, a rotation layer is added at the end of the ansatz. If True, no rotation layer is added.
• parameter_prefix – The parameterized gates require a parameter to be defined, for which we use instances of Parameter. The name of each parameter will be this specified prefix plus its index.
• insert_barriers (bool) – If True, barriers are inserted in between each layer. If False, no barriers are inserted. Defaults to False.
• initial_state – A QuantumCircuit object to prepend to the circuit.
• flatten – Set this to True to output a flat circuit instead of nesting it inside multiple layers of gate objects. By default currently the contents of the output circuit will be wrapped in nested objects for cleaner visualization. However, if you’re using this circuit for anything besides visualization its strongly recommended to set this flag to True to avoid a large performance overhead for parameter binding.

Attributes

ancillas

calibrations

clbits

data

entanglement

entanglement_blocks

extension_lib

flatten

global_phase

header

initial_state

insert_barriers

instances

layout

metadata

num_ancillas

num_clbits

num_layers

num_parameters

num_parameters_settable

num_qubits

op_start_times

ordered_parameters

>>> # prepare circuit ...
>>> print(nlocal)
     ┌───────┐┌──────────┐┌──────────┐┌──────────┐
q_0: ┤ Ry(1) ├┤ Ry(θ[1]) ├┤ Ry(θ[1]) ├┤ Ry(θ[3]) ├
     └───────┘└──────────┘└──────────┘└──────────┘
>>> nlocal.parameters
{Parameter(θ[1]), Parameter(θ[3])}
>>> nlocal.ordered_parameters
[1, Parameter(θ[1]), Parameter(θ[1]), Parameter(θ[3])]

parameter_bounds

parameters

preferred_init_points

prefix

qregs

qubits

reps

rotation_blocks