def _local_search(problem,
fringe_expander,
iterations_limit=0,
fringe_size=1,
random_initial_states=False):
'''
Basic algorithm for all local search algorithms.
'''
fringe = BoundedPriorityQueue(fringe_size)
if random_initial_states:
for _ in xrange(fringe_size):
s = problem.generate_random_state()
fringe.append(SearchNodeValueOrdered(state=s, problem=problem))
else:
fringe.append(
SearchNodeValueOrdered(state=problem.initial_state,
problem=problem))
iteration = 0
run = True
best = None
while run:
old_best = fringe[0]
fringe_expander(fringe, iteration)
best = fringe[0]
iteration += 1
if iterations_limit and iteration >= iterations_limit:
run = False
elif old_best.value() >= best.value():
run = False
return best
def _local_search(problem,
fringe_expander,
iterations_limit=0,
fringe_size=1,
random_initial_states=False,
stop_when_no_better=True,
viewer=None):
'''
Basic algorithm for all local search algorithms.
'''
if viewer:
viewer.event('started')
fringe = BoundedPriorityQueue(fringe_size)
if random_initial_states:
for _ in range(fringe_size):
s = problem.generate_random_state()
fringe.append(SearchNodeValueOrdered(state=s, problem=problem))
else:
fringe.append(
SearchNodeValueOrdered(state=problem.initial_state,
problem=problem))
finish_reason = ''
iteration = 0
run = True
best = None
while run:
if viewer:
viewer.event('new_iteration', list(fringe))
old_best = fringe[0]
fringe_expander(fringe, iteration, viewer)
best = fringe[0]
iteration += 1
if iterations_limit and iteration >= iterations_limit:
run = False
finish_reason = 'reaching iteration limit'
elif old_best.value >= best.value and stop_when_no_better:
run = False
finish_reason = 'not being able to improve solution'
if viewer:
viewer.event('finished', fringe, best,
'returned after %s' % finish_reason)
return best
def _expander(fringe, iteration, viewer):
fitness = [x.value for x in fringe]
sampler = InverseTransformSampler(fitness, fringe)
new_generation = []
expanded_nodes = []
expanded_neighbors = []
for _ in fringe:
node1 = sampler.sample()
node2 = sampler.sample()
child = problem.crossover(node1.state, node2.state)
action = 'crossover'
if random.random() < mutation_chance:
# Noooouuu! she is... he is... *IT* is a mutant!
child = problem.mutate(child)
action += '+mutation'
child_node = SearchNodeValueOrdered(state=child,
problem=problem,
action=action)
new_generation.append(child_node)
expanded_nodes.append(node1)
expanded_neighbors.append([child_node])
expanded_nodes.append(node2)
expanded_neighbors.append([child_node])
if viewer:
viewer.event('expanded', expanded_nodes, expanded_neighbors)
fringe.clear()
for node in new_generation:
fringe.append(node)
def _expander(fringe, iteration):
fitness = [x.value() for x in fringe]
sampler = InverseTransformSampler(fitness, fringe)
new_generation = []
for _ in fringe:
node1 = sampler.sample()
node2 = sampler.sample()
child = problem.crossover(node1.state, node2.state)
if random.random() < mutation_chance:
# Noooouuu! she is... he is... *IT* is a mutant!
child = problem.mutate(child)
new_generation.append(child)
fringe.clear()
for s in new_generation:
fringe.append(SearchNodeValueOrdered(state=s, problem=problem))
def test_search_node_value_sorted(self):
n1 = SearchNodeValueOrdered(problem=self.problem, state='iab')
n2 = SearchNodeValueOrdered(problem=self.problem, state='iba')
self.assertTrue(n1 < n2)
self.assertFalse(n2 < n1)