class DefaultTermination(ToleranceBasedTermination):
def __init__(self,
x_tol,
cv_tol,
f_tol,
nth_gen=5,
n_last=20,
n_max_gen=1000,
n_max_evals=100000,
**kwargs) -> None:
"""
Parameters
----------
cv_tol
x_tol
f_tol
n_max_gen : int
A limit on maximum number of generations to avoid it never terminates
(set to `None` or infinity to disable)
n_max_evals : int
A limit on maximum number of function evaluations to avoid it never terminates
(set to `None` or infinity to disable)
nth_gen
n_last
kwargs
"""
super().__init__(n_last=n_last, nth_gen=nth_gen, **kwargs)
self.n_max_gen = MaximumGenerationTermination(n_max_gen)
self.n_max_evals = MaximumFunctionCallTermination(n_max_evals)
self.x_tol = DesignSpaceToleranceTermination(tol=x_tol,
n_last=n_last,
nth_gen=nth_gen)
self.cv = ConstraintViolationToleranceTermination(tol=cv_tol,
n_last=n_last,
nth_gen=nth_gen)
self.f_tol = f_tol
def _store(self, algorithm):
return {
"n_max_gen": self.n_max_gen.do_continue(algorithm),
"n_max_evals": self.n_max_evals.do_continue(algorithm),
"x_tol": self.x_tol.do_continue(algorithm),
"cv": self.cv.do_continue(algorithm),
"f_tol": self.f_tol.do_continue(algorithm)
}
def _decide(self):
metric = self.history[-1]
return metric["n_max_gen"] and metric["n_max_evals"] and metric[
"x_tol"] and (metric["cv"] or metric["f_tol"])
def __init__(self,
x_tol,
cv_tol,
f_tol,
nth_gen=5,
n_last=20,
n_max_gen=1000,
n_max_evals=100000,
**kwargs) -> None:
"""
Parameters
----------
cv_tol
x_tol
f_tol
n_max_gen : int
A limit on maximum number of generations to avoid it never terminates
(set to `None` or infinity to disable)
n_max_evals : int
A limit on maximum number of function evaluations to avoid it never terminates
(set to `None` or infinity to disable)
nth_gen
n_last
kwargs
"""
super().__init__(n_last=n_last, nth_gen=nth_gen, **kwargs)
self.n_max_gen = MaximumGenerationTermination(n_max_gen)
self.n_max_evals = MaximumFunctionCallTermination(n_max_evals)
self.x_tol = DesignSpaceToleranceTermination(tol=x_tol,
n_last=n_last,
nth_gen=nth_gen)
self.cv = ConstraintViolationToleranceTermination(tol=cv_tol,
n_last=n_last,
nth_gen=nth_gen)
self.f_tol = f_tol
def setup(self, problem, **kwargs):
super().setup(problem, **kwargs)
# if maximum functions termination convert it to generations
if isinstance(self.termination, MaximumFunctionCallTermination):
n_gen = np.ceil((self.termination.n_max_evals - self.pop_size) / self.n_offsprings)
self.termination = MaximumGenerationTermination(n_gen)
# check whether the n_gen termination is used - otherwise this algorithm can be not run
if not isinstance(self.termination, MaximumGenerationTermination):
raise Exception("Please use the n_gen or n_eval as a termination criterion to execute RVEA!")
return self
def _setup(self, problem, **kwargs):
self.norm = ObjectiveSpaceNormalization()
# if maximum functions termination convert it to generations
if isinstance(self.termination, MaximumFunctionCallTermination):
n_gen = np.ceil((self.termination.n_max_evals - self.pop_size) /
self.n_offsprings)
self.termination = MaximumGenerationTermination(n_gen)
# check whether the n_gen termination is used - otherwise this algorithm can be not run
if not isinstance(self.termination, MaximumGenerationTermination):
raise Exception(
"Please use the n_gen or n_eval as a termination criterion to run ICMA!"
)
def __init__(self,
metric_window_size=None,
data_window_size=None,
min_data_for_metric=1,
nth_gen=1,
n_max_gen=None,
n_max_evals=None,
truncate_metrics=True,
truncate_data=True,
) -> None:
"""
Parameters
----------
metric_window_size : int
The last generations that should be considering during the calculations
data_window_size : int
How much of the history should be kept in memory based on a sliding window.
nth_gen : int
Each n-th generation the termination should be checked for
"""
super().__init__(MaximumGenerationTermination(n_max_gen=n_max_gen),
MaximumFunctionCallTermination(n_max_evals=n_max_evals))
# the window sizes stored in objects
self.data_window_size = data_window_size
self.metric_window_size = metric_window_size
# the obtained data at each iteration
self.data = SlidingWindow(data_window_size) if truncate_data else []
# the metrics calculated also in a sliding window
self.metrics = SlidingWindow(metric_window_size) if truncate_metrics else []
# each n-th generation the termination decides whether to terminate or not
self.nth_gen = nth_gen
# number of entries of data need to be stored to calculate the metric at all
self.min_data_for_metric = min_data_for_metric
def from_args(cls, args: Namespace, index=None) -> Termination:
n = cls._parsed_argument('n', args, index=index)
return MaximumGenerationTermination(n)