Using Ansätze

Ansätze are pre-built quantum circuit patterns that serve as templates for quantum algorithms. Out HSMPO surrogate works around a provided ansatz, but, as you know from previous examples, a simple Qibo circuit works too. This is because we internally take care of constructing a proper ansatz around the provided circuit.

Available Ansätze

In MPSTAB, we provide a few precomputed ansätze, for example Hardware Efficient or Floquet Dynamics inspired. More remarkably, we provide two, more abstract, models: a first general wrapper, which can be used to transform any Qibo circuit into an Mpstab ansatz (CircuitAnsatz), and a more complicated ansatz, which allows to provide more information about the quantum hardware and, then, recover a compiled and transpiled ansatz (TranspiledAnsatz). This second model is particularly relevant in contexts where one wants to have control of the number of gates executed on a real quantum device.

CircuitAnsatz

Wrap any existing Qibo circuit as an ansatz:

from mpstab import HSMPO
from qibo import Circuit, gates
from mpstab.models.ansatze import CircuitAnsatz

# Create a custom circuit
circuit = Circuit(5)
for q in range(5):
    circuit.add(gates.H(q))
    circuit.add(gates.RY(q, theta=0.5))
circuit.add(gates.CNOT(0, 1))

# Wrap as ansatz
ansatz = CircuitAnsatz(qibo_circuit=circuit)
simulator = HSMPO(ansatz=ansatz)

Building Custom Ansätze

Extend the Ansatz class for custom patterns:

from mpstab import HSMPO
from mpstab.models.ansatze import Ansatz
from qibo import Circuit, gates
import numpy as np

class MyCustomAnsatz(Ansatz):
    """Custom ansatz for specific problem."""

    def __init__(self, nqubits, pattern="ring"):
        super().__init__(nqubits)
        self.pattern = pattern
        self._build_circuit()

    def _build_circuit(self):
        # Your circuit construction logic
        self.circuit = Circuit(self.nqubits)

        # Add gates based on pattern
        if self.pattern == "ring":
            for q in range(self.nqubits):
                self.circuit.add(gates.H(q))
            for q in range(self.nqubits):
                next_q = (q + 1) % self.nqubits
                self.circuit.add(gates.CNOT(q, next_q))

# Use with HSMPO
ansatz = MyCustomAnsatz(nqubits=5, pattern="ring")
simulator = HSMPO(ansatz=ansatz)

The ansatz partitioning

An HSMPO surrogate is constructed grouping all the Clifford operations in the circuit and rewriting the representation of the unitary in terms of new - global - rotations. All this procedure, that is well described in the original HSMPO paper, requires a fine partitioning of the original quantum circuit, and a reconstruction of a series of operations alternately handled by the stabilizers and the tensor network engines. Our ansätze provide a built int method to partitionate and recover all these features of the provided circuit.

Further Reading