def __init__(self, domain, ce_params):
super().__init__(domain)
self.alpha = ce_params.alpha
self.thres = ce_params.thres
self.cont_buckets = ce_params.cont.buckets
self.cont_dist = ce_params.cont.dist
self.disc_dist = ce_params.disc.dist
self.cont_ce = lambda domain: ContinuousCrossEntropySampler(
domain=domain,
buckets=self.cont_buckets,
dist=self.cont_dist,
alpha=self.alpha,
thres=self.thres)
self.disc_ce = lambda domain: DiscreteCrossEntropySampler(
domain=domain,
dist=self.disc_dist,
alpha=self.alpha,
thres=self.thres)
partition = ((lambda d: d.standardizedDimension > 0, self.cont_ce),
(lambda d: d.standardizedIntervals, self.disc_ce))
self.split_sampler = SplitSampler.fromPartition(
domain, partition, RandomSampler)
self.cont_sampler, self.disc_sampler = None, None
self.rand_sampler = None
for subsampler in self.split_sampler.samplers:
if isinstance(subsampler, ContinuousCrossEntropySampler):
assert self.cont_sampler is None
self.cont_sampler = subsampler
elif isinstance(subsampler, DiscreteCrossEntropySampler):
assert self.disc_sampler is None
self.disc_sampler = subsampler
else:
assert isinstance(subsampler, RandomSampler)
assert self.rand_sampler is None
self.rand_sampler = subsampler
def makeDomainSampler(domain):
return SplitSampler.fromPredicate(
domain,
lambda d: d.standardizedDimension > 0,
lambda domain: SimulatedAnnealingSampler(domain=domain,
sa_params=sa_params),
makeRandomSampler)
def makeDomainSampler(domain):
return SplitSampler.fromPredicate(
domain,
lambda d: d.isStandardizable,
lambda domain: GridSampler(domain=domain,
grid_params=grid_params),
makeRandomSampler)
def makeDomainSampler(domain):
return SplitSampler.fromPredicate(
domain,
lambda d: d.standardizedDimension > 0,
lambda domain: HaltonSampler(domain=domain,
halton_params=halton_params),
makeRandomSampler)
def makeDomainSampler(domain):
return SplitSampler.fromPredicate(
domain,
lambda d: d.standardizedDimension > 0,
lambda domain: BayesOptSampler(domain=domain,
BO_params=BO_params),
makeRandomSampler)
def __init__(self, domain, ce_params):
super().__init__(domain)
self.f = ce_params.f
self.alpha = ce_params.alpha
self.thres = ce_params.thres
self.cont_buckets = ce_params.cont.buckets
self.cont_dist = ce_params.cont.dist
self.disc_dist = ce_params.disc.dist
self.cont_ce = lambda domain: ContinuousCrossEntropySampler(
domain=domain,
buckets=self.cont_buckets,
dist=self.cont_dist,
alpha=self.alpha)
self.disc_ce = lambda domain: DiscreteCrossEntropySampler(
domain=domain, dist=self.disc_dist, alpha=self.alpha)
partition = ((lambda d: d.standardizedDimension, self.cont_ce),
(lambda d: d.standardizedIntervals, self.disc_ce))
self.split_sampler = SplitSampler.fromPartition(
domain, partition, RandomSampler)
def __init__(self, domain, grid_params=None):
if grid_params is None:
grid_params = {}
self.cont_N = grid_params.get('N', 21)
repeat = grid_params.get('repeat', False)
super().__init__(domain, repeat=repeat)
cont_grid = lambda domain: ContinuousGridSampler(domain=domain,
N=self.cont_N)
disc_grid = lambda domain: DiscreteGridSampler(domain=domain)
partition = (
(lambda d: d.standardizedDimension >= 0, cont_grid),
(lambda d: d.standardizedIntervals, disc_grid)
)
self.split_sampler = SplitSampler.fromPartition(domain, partition)
self.cont_sampler, self.disc_sampler = \
self.split_sampler.samplersForPredicates