def evolve(self):
new_population = []
new_scores = defaultdict(int)
mutation_p = 0.01
for x in xrange(self.population_size - 40):
first = choice(self.classifiers)
second = choice(self.classifiers)
if self.scores[first] >= self.scores[second]:
new_population.append(first)
new_scores[first] = self.scores[first]
else:
new_population.append(second)
new_scores[second] = self.scores[second]
def crossover(first, second):
idx = choice(xrange(len(first.condition)))
condition = first[:idx] + second[idx:]
action = choice([first.action, second.action])
return Classifier(condition=condition, action=action)
crossovers = []
for x in xrange(self.population_size - len(new_population)):
first = choice(new_population)
second = choice(new_population)
crossovers.append(crossover(first, second))
def mutate_cond(clf):
idx = choice(xrange(len(clf.condition)))
condition = str(clf.condition[:idx]) + choice(
["0", "1", "#"]) + str(clf.condition[idx + 1:])
return Classifier(condition=condition, action=clf.action)
def mutate_action(clf):
action = choice(self.actions)
return Classifier(condition=clf.condition, action=action)
mutants = []
for clf in new_population:
if random_f() < mutation_p:
new_clf = mutate_cond(clf)
mutants.append(new_clf)
for clf in new_population:
if random_f() < mutation_p:
new_clf = mutate_action(clf)
mutants.append(new_clf)
self.classifiers = new_population + crossovers + mutants
self.scores = new_scores
def evolve(self):
new_population = []
new_scores = defaultdict(int)
mutation_p = 0.01
for x in xrange(self.population_size - 40):
first = choice(self.classifiers)
second = choice(self.classifiers)
if self.scores[first] >= self.scores[second]:
new_population.append(first)
new_scores[first] = self.scores[first]
else:
new_population.append(second)
new_scores[second] = self.scores[second]
def crossover(first, second):
idx = choice(xrange(len(first.condition)))
condition = first[:idx] + second[idx:]
action = choice([first.action, second.action])
return Classifier(condition=condition, action=action)
crossovers = []
for x in xrange(self.population_size - len(new_population)):
first = choice(new_population)
second = choice(new_population)
crossovers.append(crossover(first, second))
def mutate_cond(clf):
idx = choice(xrange(len(clf.condition)))
condition = str(clf.condition[:idx]) + choice(["0", "1", "#"]) + str(clf.condition[idx + 1 :])
return Classifier(condition=condition, action=clf.action)
def mutate_action(clf):
action = choice(self.actions)
return Classifier(condition=clf.condition, action=action)
mutants = []
for clf in new_population:
if random_f() < mutation_p:
new_clf = mutate_cond(clf)
mutants.append(new_clf)
for clf in new_population:
if random_f() < mutation_p:
new_clf = mutate_action(clf)
mutants.append(new_clf)
self.classifiers = new_population + crossovers + mutants
self.scores = new_scores
def __init__(self, secs=None, random=None, debug=False):
self.debug=debug
if secs:
secs=float(secs)
self.dwell=lambda : secs
else:
rand_spec=random
rand_spec_doc="""
"scale:offset"
scale
(scale,offset)
"""
if isinstance(rand_spec, basestring):
scale,offset=map(float,rand_spec.split(':'))
elif isinstance(rand_spec, (list,tuple)):
scale,offset=rand_spec
elif isinstance(rand_spec, (int,float)):
scale,offset=rand_spec,0
else:
raise RuntimeError("invalid random sleep specification. 'random' must be one of:\n" \
+ rand_spec_doc)
self.dwell=lambda : scale*random_f()+offset