def __init__(self,
device,
name,
tolerance,
t_max,
local_search_iterations=0,
epsilon=0.05):
super(SurveyPropagatorSolver, self).__init__(
device=device,
name=name,
propagator=pdp_propagate.SurveyPropagator(device,
decimator_dimension=1,
include_adaptors=False),
decimator=pdp_decimate.SequentialDecimator(
device,
message_dimension=(3, 1),
scorer=pdp_predict.SurveyScorer(device,
message_dimension=1,
include_adaptors=False),
tolerance=tolerance,
t_max=t_max),
predictor=pdp_predict.IdentityPredictor(device=device,
random_fill=True),
local_search_iterations=local_search_iterations,
epsilon=epsilon)
def __init__(
self,
device,
name,
edge_dimension,
meta_data_dimension,
propagator_dimension,
decimator_dimension,
mem_hidden_dimension,
agg_hidden_dimension,
mem_agg_hidden_dimension,
classifier_dimension,
dropout,
tolerance,
t_max,
local_search_iterations=0,
epsilon=0.05,
):
super(NeuralSequentialDecimatorSolver, self).__init__(
device=device,
name=name,
propagator=pdp_propagate.NeuralMessagePasser(
device,
edge_dimension,
decimator_dimension,
meta_data_dimension,
propagator_dimension,
mem_hidden_dimension,
mem_agg_hidden_dimension,
agg_hidden_dimension,
dropout,
),
decimator=pdp_decimate.SequentialDecimator(
device,
message_dimension=(3, 1),
scorer=pdp_predict.NeuralPredictor(
device,
decimator_dimension,
1,
edge_dimension,
meta_data_dimension,
mem_hidden_dimension,
agg_hidden_dimension,
mem_agg_hidden_dimension,
variable_classifier=util.PerceptronTanh(
decimator_dimension, classifier_dimension, 1
),
function_classifier=None,
),
tolerance=tolerance,
t_max=t_max,
),
predictor=pdp_predict.IdentityPredictor(device=device, random_fill=True),
local_search_iterations=local_search_iterations,
epsilon=epsilon,
)
def __init__(self, device, name, iteration_num, epsilon=0.05):
super(WalkSATSolver, self).__init__(
device=device,
name=name,
propagator=None,
decimator=None,
predictor=pdp_predict.IdentityPredictor(device=device, random_fill=True),
local_search_iterations=iteration_num,
epsilon=epsilon,
)