def CreatePenaltyEvaluator(config):
penEval = EA.PenaltyEvaluator()
objectiveSpecs = config["Objectives"]
for spec in objectiveSpecs:
name = spec["Name"]
penaltyType = spec["Type"]
weight = spec["Weight"]
ref = spec["Reference"]
exponent = spec["Exponent"]
penalty = EA.Penalty(penaltyType, name, weight, ref, exponent=exponent)
penEval.ObjectivePenalties.append(penalty)
if not config.has_key("Constraints"):
return penEval
constraintSpecs = config["Constraints"]
for spec in constraintSpecs:
name = spec["Name"]
penaltyType = spec["Type"]
weight = spec["Weight"]
ref = spec["Reference"]
imposed = spec["Imposed"]
exponent = spec["Exponent"]
penalty = EA.Penalty(penaltyType,
name,
weight,
ref,
imposed=imposed,
exponent=exponent)
penEval.ConstraintPenalties.append(penalty)
return penEval
def Start(self): pop = EA.Population() for i in range(5): indi = self.Chromosome.Clone() pop.append(indi) print pop self.ResEval.Evaluate(pop)
def __init__(self, spec):
Method.__init__(self)
self.Chromosome = Factory.Chromosome(Context.SimulationChain)
self.ResEval = ChainResponseEvaluator()
self.ResEval.Start()
self.PenEval = Factory.CreatePenaltyEvaluator(spec)
self.FitnessEval = EA.FitnessEvaluator()
optiConfig = spec["OptimizerConfig"]
popSize = optiConfig["PopulationSize"]
self.OptimizerConfig = optiConfig
postIterCallback = lambda gen, pop: self.PostIteration(gen, pop)
self.BestExperiment = None
if spec["Optimizer"] == "GA":
self.Optimizer = EA.MonoObjectiveGA(self.Chromosome, self.ResEval,
self.PenEval, self.FitnessEval,
popSize, 2)
self.Optimizer.PostIterationCallback = postIterCallback
self.HistoryFile = open("%s.history.dat" % Context.Name, "w")
elif spec["Optimizer"] == "SurrogateGA":
Context.Database.Import(open(optiConfig["DOE"]))
pop = Factory.PopulationFromPlan(self.Chromosome, Context.Database)
self.Optimizer = EA.MonoObjectiveSurrogateGA(
self.Chromosome, self.ResEval, self.PenEval, self.FitnessEval,
pop)
elif spec["Optimizer"] == "ParetoGA":
self.Optimizer = EA.SPEA2(self.Chromosome, self.ResEval,
self.PenEval,
optiConfig["PopulationSize"],
optiConfig["ArchiveSize"])
elif spec["Optimizer"] == "ParetoSurrogateGA":
Context.Database.Import(open(optiConfig["DOE"]))
pop = Factory.PopulationFromPlan(self.Chromosome, Context.Database)
self.Optimizer = EA.ParetoSurrogateGA(
self.Chromosome, self.ResEval, self.PenEval, pop, popSize,
optiConfig["ArchiveSize"], optiConfig["Update"],
optiConfig["UpdateTruncation"])
def Chromosome(chain):
indi = EA.Individual()
for param in chain.Parameters:
name = param["Name"]
valueType = param["Type"]
minValue = param["Min"]
maxValue = param["Max"]
refValue = param["Ref"]
p = EA.BoundedValue(name, valueType, refValue, refValue, minValue,
maxValue)
indi.DesignParameters.append(p)
for resp in chain.Responses:
name = resp["Name"]
valueType = resp["Type"]
r = EA.Value(name, valueType, 0.0)
indi.Responses.append(r)
return indi
Rules.fill_dict(users, train_user_label)
train_user_collect = Rules.load_dict("train/topk2.txt")
Rules.fill_dict(users, train_user_collect)
train_item_reco = Rules.load_dict("train/item_based_user_dict.txt")
train_user_reco = Rules.Similarity(train_similarity, train_user_item,
train_topk4)
train_item_action = Rules.load_dict("train/user_action.txt")
Rules.fill_dict(users, train_item_action)
train_user_matrix, train_pool_dict, train_right_dict = Rules.classification(
train_notbuy, train_hottest, train_topview, train_user_reco,
train_item_reco, users, train_user_label, train_user_item,
train_item_action)
train_weights = EA.calculateWeights(train_hottest, train_notbuy,
train_topview, train_item_reco,
train_user_reco, train_user_label)
train_result = EA.EArecommend(train_hottest, train_notbuy, train_topview,
train_item_reco, train_user_reco,
train_weights)
EA.judge(train_result, train_user_label)
train_similarity = {}
train_user_item = {}
train_topk4 = {}
train_notbuy = {}
train_hottest = {}
train_topview = {}
train_user_reco = {}
train_item_reco = {}
#train_user_label = {}
# $Id: TestTorczonMeritFunction.py 87 2011-02-17 16:46:52Z kato $
import EA
def Main():
case = EA.TestCases.Rosenbrock()
chromosome = case.Chromosome()
chromosome.Responses.append(EA.Value("TorczonMerit", "FLOAT"))
resEval = EA.ResponseEvaluator(case)
penaltyEval = EA.PenaltyEvaluator()
penaltyEval.ObjectivePenalties = case.Objectives()
penaltyEval.ConstraintPenalties = case.Constraints()
fitnessEval = EA.FitnessEvaluator()
print chromosome
pop = EA.Population()
for i in range(10):
pop.append(chromosome.Clone())
pop.Randomize()
resEval.Evaluate(pop)
penaltyEval.Evaluate(pop)
fitnessEval.Evaluate(pop)
torczon = EA.TorczonMeritFunction(pop, chromosome, 1.0, "TorczonMerit")
def ind_test():
board = ea.erzeuge_individuum(numero)
assert len(board.flat) == numero * numero
def PopulationFromPlan(chromosome, plan):
pop = EA.Population()
for exp in plan:
pop.append(IndividualFromExperiment(chromosome, exp))
return pop
def main():
# pass
ea = EA(POP_SIZE, GEN_NUM, GENE_LEN)
best_window_size, best_num_units = ea.evo_algorithm()
model, test_acc = train_model(best_window_size, best_num_units)
# $Id: TestDebConstraintFitness.py 87 2011-02-17 16:46:52Z kato $
import EA
def Main():
case = EA.TestCases.RosenbrockUncomputable()
resEval = EA.ResponseEvaluator(case)
penaltyEval = EA.PenaltyEvaluator()
penaltyEval.ObjectivePenalties = case.Objectives()
penaltyEval.ConstraintPenalties = case.Constraints()
fitnessEval = EA.FitnessEvaluator()
pop = EA.Population()
for i in range(10):
pop.append(case.Chromosome())
pop.Randomize()
resEval.Evaluate(pop)
penaltyEval.Evaluate(pop)
print pop
fitnessEval.Evaluate(pop)
pop.SortByFitness()
print pop