Simulated Annealing

`SimulatedAnnealingSolver`

Simple classical solver class using Simulated Annealing. Designed to integrate with the solver factory.

Signature

class SimulatedAnnealingSolver(BaseClassicalSolver):
    def solve(self) -> QUBOSolution

Description

This solver uses a Simulated Annealing to probabilistically explore the solution space. It is suitable for approximating solutions on medium-sized QUBO instances. Computation is entirely classical and based on the SimulatedAnnealingSolver. The output is fully compatible with the QUBOSolution structure used in the qubo-solver package.

Fields

Field	Type	Description
`use_quantum`	`bool`	Have to be `False` to uses a classical solver.
`classical_solver_type`	`str`	Set to `"simulated_annealing"` to use Simulated Annealing as the solving method.
`max_iter`	`int`	Maximum number of iterations to perform for simulated annealing or tabu search.
`sa_initial_temp`	`float`	Starting temperature (controls exploration).
`sa_final_temp`	`float`	Minimum temperature threshold for stopping.
`sa_alpha`	`float`	Cooling rate - should be slightly below 1 (e.g., 0.95–0.99).

Usage

from qubosolver import QUBOInstance
from qubosolver.solver import QuboSolver
from qubosolver.config import SolverConfig, ClassicalConfig

qubo = QUBOInstance(coefficients=[[-2.0, 1.0], [1.0, -2.0]])
config = SolverConfig(use_quantum = False, classical=ClassicalConfig(classical_solver_type="simulated_annealing"))

solver = QuboSolver(qubo, config)

solution = solver.solve()
print(solution)

QUBOSolution(bitstrings=tensor([[1., 1.]]), costs=tensor([-2.]), counts=None, probabilities=None, solution_status=)

Notes

Recommended for local, classical solving when exact optimization is not required.