qilisdk.utils.classical_solvers.simulated_annealing_solver

Classes

SimulatedAnnealingSolver

Classical solver that uses simulated annealing, implemented in C++.

Module Contents

class SimulatedAnnealingSolver(num_reads: int = 10, num_sweeps: int = 1000, beta_range: tuple[float, float] | None = None, seed: int = 0, num_threads: int = 0)

Bases: qilisdk.utils.classical_solvers.base_solver.ClassicalSolver

Classical solver that uses simulated annealing, implemented in C++. This solves a QUBO model and rejects others.

Example

from qilisdk.core import Model
from qilisdk.utils.classical_solvers import SimulatedAnnealingSolver

model = Model.knapsack(values=[5, 4], weights=[3, 2], max_weight=3)
result = SimulatedAnnealingSolver(num_reads=100).solve(model.to_qubo())

Create a new simulated annealing based classical solver instance.

Parameters:

• num_reads (int, optional) – The number of independent anneals to run, the best of which is returned. Defaults to 10.

• num_sweeps (int, optional) – The number of sweeps over all variables in each anneal. Defaults to 1000.

• beta_range (tuple[float, float] | None, optional) – The (initial, final) inverse temperature to anneal over. If not given, a range is derived from the magnitudes of the cost function’s coefficients. Defaults to None.

• seed (int, optional) – The seed of the random number generators, each read deriving its own from it. Defaults to 0.

• num_threads (int, optional) – The number of threads to distribute the reads over, or zero to let OpenMP decide. Defaults to 0.

num_reads = 10

num_sweeps = 1000

beta_range = None

seed = 0

num_threads = 0

solve(model: qilisdk.core.Model) → qilisdk.utils.classical_solvers.base_solver.ClassicalSolverResult

Solve the given QUBO by annealing it in C++.

Parameters:

model – The QUBO instance to solve. Typed as Model to keep the ClassicalSolver interface, but anything other than a QUBO is rejected, so a general Model must be converted with to_qubo() first.

Returns:

the results of the optimization, including the objective value and best solution.

Return type:

ClassicalSolverResult

Raises:

ValueError – if the given model is not a QUBO, or if the annealing settings are invalid.