def sample(self, bqm, num_repeats=50, seed=None, algorithm=None,
verbosity=-1, timeout=2592000, solver_limit=None, solver=None,
target=None, find_max=False, **sample_kwargs):
"""Sample low-energy states defined by a QUBO using qbsolv.
Note:
The qbsolv library being shared by all instances of this class is
non-reentrant and not thread safe. The GIL should not be released
by this method until that is resolved.
Note:
The default build of this library doesn't have the dw library.
To use solver='dw' this module must be built from source with
that library.
The parameter `solver` given to this method has several valid forms:
- String 'tabu' (default): sub problems are called via an internal call to tabu.
- String 'dw': sub problems are given to the dw library.
- Instance of a dimod sampler. The `sample_qubo` method is invoked.
- Callable that has the signature (qubo: dict, current_best: dict)
and returns a result list/dictionary with the new solution.
Args:
Q (dict): A dictionary defining the QUBO. Should be of the form
{(u, v): bias} where u, v are variables and bias is numeric.
num_repeats (int, optional): Determines the number of times to
repeat the main loop in qbsolv after determining a better
sample. Default 50.
seed (int, optional): Random seed. Default generated by random module.
algorithm (int, optional): Algorithm to use. Default is
ENERGY_IMPACT. Algorithm numbers can be imported from the module
under the names ENERGY_IMPACT and SOLUTION_DIVERSITY.
verbosity (int, optional): Prints more detail about qbsolv's internal
process as this number increases.
timeout (float, optional): Number of seconds before routine halts. Default is 2592000.
solver: Sampling method for qbsolv to use; see method description.
solver_limit (int, optional): Maximum number of variables in a sub problem.
target (float, optional): If given, qbsolv halts when
a state with this energy value or better is discoverd. Default is None.
find_max (bool, optional): Switches from searching for minimization to
maximization. Default is False (minimization).
Returns:
:obj:`Response`
Examples:
This example uses the tabu search algorithm to solve a small QUBO.
>>> Q = {(0, 0): 1, (1, 1): 1, (0, 1): 1}
>>> response = QBSolv().sample_qubo(Q)
>>> list(response.samples())
'[{0: 0, 1: 0}]'
>>> list(response.energies())
'[0.0]'
"""
if not isinstance(num_repeats, int) or num_repeats <= 0:
raise ValueError("num_repeats must be a positive integer")
# pose the QUBO to qbsolv
Q, offset = bqm.to_qubo()
samples, energies, counts = run_qbsolv(Q=Q, num_repeats=num_repeats, seed=seed, algorithm=algorithm,
verbosity=verbosity, timeout=timeout, solver_limit=solver_limit,
solver=solver, target=target, find_max=find_max, sample_kwargs=sample_kwargs)
response = dimod.SampleSet.from_samples(samples, energy=energies,
num_occurrences=counts, vartype=dimod.BINARY)
response.change_vartype(bqm.vartype, energy_offset=offset)
return response
def sample_qubo(self,
Q,
num_repeats=50,
seed=None,
algorithm=None,
verbosity=-1,
timeout=2592000,
solver_limit=None,
solver=None,
target=None,
find_max=False):
"""Sample low-energy states defined by a QUBO using qbsolv.
Note:
The qbsolv library being shared by all instances of this class is
non-rentrant and not thread safe. The GIL should not be released
by this method until that is resolved.
Note:
The default build of this library doesn't have the dw library,
to use solver='dw' this module must be built from source with
that library.
The parameter solver given to this method has several valid forms:
- (default) The string 'tabu': sub problems will be called via an internal call to tabu.
- The string 'dw': sub problems will be given to the dw library.
- an instance of a dimod sampler, the sample_qubo method will be invoked.
- an callable that has the signature (qubo: dict, current_best: dict)
and returns a result list/dictionary with the new solution
Args:
Q (dict): A dictionary defining the QUBO. Should be of the form
{(u, v): bias} where u, v are variables and bias is numeric.
num_repeats (int, optional): Determines the number of times to
repeat the main loop in qbsolv after determining a better
sample. Default 50.
seed (int, optional): Random seed. Default generated by random module.
algorithm (int, optional): Which algorithm to use. Defaut
ENERGY_IMPACT. Algorithms numbers can be imported from module
under the names ENERGY_IMPACT and SOLUTION_DIVERSITY.
verbosity (int, optional): Print more detail about qbsolv's internal
process as this number gets higher.
timeout (float, optional): Number of seconds before routine halts. Default: 2592000
solver: The sampling method for qbsolv to use, see method description.
solver_limit (int, optional): The maximum number of variables in a sub problem.
target (float, optional): If given the qbsolv process haults when
a state with this energy value or better is discoverd. (default None)
find_max (bool, optional): Switches the process from minimization to
maximuzation. (default False, minimization)
Returns:
:obj:`BinaryResponse`
Examples:
>>> Q = {(0, 0): 1, (1, 1): 1, (0, 1): 1}
>>> response = QBSolv().sample_qubo(Q)
>>> list(response.samples())
'[{0: 0, 1: 0}]'
>>> list(response.energies())
'[0.0]'
"""
if not isinstance(num_repeats, int) or num_repeats <= 0:
raise ValueError("num_repeats must be a positive integer")
# pose the QUBO to qbsolv
samples, counts = run_qbsolv(Q=Q,
num_repeats=num_repeats,
seed=seed,
algorithm=algorithm,
verbosity=verbosity,
timeout=timeout,
solver_limit=solver_limit,
solver=solver,
target=target,
find_max=find_max)
# load the response
response = dimod.BinaryResponse()
response.add_samples_from(samples,
sample_data=({
'count': n
} for n in counts),
Q=Q)
return response