def avm(self, methods, inputs=None):
for method in methods:
for j in range(self.attempts):
# print("Starting inputs, iteration ", inputs, j)
if not inputs:
inputs = [
random.randint(-self.range, self.range)
for i in range(self.arg_count)
]
# start_inputs = copy.deepcopy(inputs)
for i, value in enumerate(inputs):
try:
value, fitness = method(inputs, i)
except AnswerFound:
# print("Found Answer")
# print(self.answer)
return self.answer
except TimeExceeded:
return "Unable to find solution, time exceeded", inputs
if fitness <= 0.0 and self.satisfied_condition(inputs):
# break
return inputs, calculate_fitness(
self.tree, inputs, self.path, self.func_name)
inputs[i] = value
if self.satisfied_condition(inputs):
return inputs, calculate_fitness(self.tree, inputs,
self.path, self.func_name)
inputs = None
self.range *= 10
# print(self.results)
# print('---')
# for res in self.results.items():
# print(res)
# print('---')
return "Unable to find solution", inputs
def mutate_population(population, scores):
for i in range(len(population)):
if uniform(0, 1) < MUTATION:
population[i] = mutate_conference(population[i])
scores[i] = calculate_fitness(population[i])
return (population, scores)
def get_f(self, inputs, index, value):
inputs[index] = value
if str(inputs) in self.results:
return self.results[str(inputs)]
else:
fitness, branch_state, approach_level = calculate_fitness(
self.tree, inputs, self.path, self.func_name)
# self.iterations += 1
# # print(self.iterations)
# if self.iteration_limit < self.iterations:
# print("Iterations exceeded")
# raise IterationsExceeded
# if not self.state:
# print(fitness, branch_state, inputs)
if approach_level == 0 and branch_state == self.state:
self.answer = inputs, (fitness, branch_state, approach_level)
raise AnswerFound
self.results[str(inputs)] = fitness
return fitness
def find_next_move(self):
grid_backup = copy.deepcopy(self.grid)
block_backup = self.block.copy()
position_backup = self.position
# TODO find an actual good min here
best_score = -999
best_x = 0
best_rotation = 0
for num_rotation in range(4):
self.move_block((0, 0))
if num_rotation > 0:
self.rotate()
while self.move_left():
pass
for x in range(self.width):
self.flash()
fitness_score = calculate_fitness(self.grid, self.trait_sets)
if fitness_score > best_score:
best_score = fitness_score
best_x = self.position[0]
best_rotation = num_rotation
self.move_block((self.position[0], 0))
if self.move_right() == False:
break
self.hide_current_block()
self.position = position_backup
self.block = block_backup
self.grid = grid_backup
return (best_x, best_rotation)
def get_fitness(self,cost): return fit.calculate_fitness(self.pos,cost)
sequence=np.concatenate((sequence,self.vel[0:int(c1)])) #if c2>=1: # sequence=np.concatenate((sequence,seq1)) # c2=c2-1 sequence=np.concatenate((sequence,seq1[0:int(c2)])) #if c3>=1: # sequence=np.concatenate((sequence,seq2)) # c3=c3-1 sequence=np.concatenate((sequence,seq2[0:int(c3)])) return sequence base=np.arange(cities) base=base+1 base=np.random.permutation(base) gbest_fit=fit.calculate_fitness(base,cost) pop=[] flag=0 #Initialize the particle positions,velocities,fitness for i in range(0,no_of_particles): temp1=np.random.permutation(base) temp2=np.random.permutation(base) fit_val1=fit.calculate_fitness(temp1,cost) fit_val2=fit.calculate_fitness(temp2,cost) if fit_val1<=fit_val2: #sequence=swap.get_swap_sequence(temp1,np.copy(temp2)) #sequence=np.random.permutation(sequence)
def test_fitness():
config = generate_nginx_config.generate_random_config()
this_fitness = fitness.calculate_fitness(config)
assert (this_fitness)
assert (this_fitness == 999)
for e in tree.edges:
if random.random() < 0.3:
return e
else:
return get_node(e)
return tree
return tree
def crossover(father, mother):
node = get_node(mother)
father.append(node)
return mother
fits = []
for t in trees:
fits.append(calculate_fitness(t))
fits.sort()
print fits[:5]
for i in range(60):
fits = {}
for t in trees:
fit = calculate_fitness(t)
fits[fit] = t
keys = fits.keys()
keys.sort()
keys = keys[:int(GENERATION_SIZE / 2)]
trees = []
for t in keys:
for i in range(2):
def satisfied_condition(self, inputs):
return calculate_fitness(self.tree, inputs, self.path,
self.func_name)[1] == self.state
def get_fitness(self, cost):
return fit.calculate_fitness(self.pos, cost)
sequence = np.concatenate((sequence, self.vel[0:int(c1)]))
#if c2>=1:
# sequence=np.concatenate((sequence,seq1))
# c2=c2-1
sequence = np.concatenate((sequence, seq1[0:int(c2)]))
#if c3>=1:
# sequence=np.concatenate((sequence,seq2))
# c3=c3-1
sequence = np.concatenate((sequence, seq2[0:int(c3)]))
return sequence
base = np.arange(cities)
base = base + 1
base = np.random.permutation(base)
gbest_fit = fit.calculate_fitness(base, cost)
pop = []
flag = 0
#Initialize the particle positions,velocities,fitness
for i in range(0, no_of_particles):
temp1 = np.random.permutation(base)
temp2 = np.random.permutation(base)
fit_val1 = fit.calculate_fitness(temp1, cost)
fit_val2 = fit.calculate_fitness(temp2, cost)
if fit_val1 <= fit_val2:
#sequence=swap.get_swap_sequence(temp1,np.copy(temp2))
#sequence=np.random.permutation(sequence)
sequence = swap.get_random_sequence(cities)
if len(sequence) != 0:
temp_pop=np.zeros(shape=(pop_size,cities))
fval=np.zeros(shape=(pop_size,2))
counter=0 #variable used to terminate when convergence occurs
#cost=np.loadtxt('gr17_d.txt')
cost=np.loadtxt('dantzig42_d.txt')
#initializing the population
for i in range(0,pop_size):
pop[i]=np.random.permutation(base)
#print pop[i]
#calculate the fitness of all individuals in the population
for i in range(0,pop_size):
fval[i][0]=i
fval[i][1]=fit.calculate_fitness(pop[i],cost)
#print fval[i][1]
initial_max_fit=min(fval[:, 1])
#print pop
print initial_max_fit
for k in range(0,generations):
new_pop=np.zeros(shape=(pop_size,cities))
#print 'hello'
#select parents from matting pool with tournament selection
for i in range(0,pop_size):
indx1=random.randrange(0,pop_size)
indx2=random.randrange(0,pop_size)
fit_val1=fit.calculate_fitness(pop[indx1],cost)
pop = np.zeros(shape=(pop_size, cities))
temp_pop = np.zeros(shape=(pop_size, cities))
fval = np.zeros(shape=(pop_size, 2))
counter = 0 #variable used to terminate when convergence occurs
#cost=np.loadtxt('gr17_d.txt')
cost = np.loadtxt('dantzig42_d.txt')
#initializing the population
for i in range(0, pop_size):
pop[i] = np.random.permutation(base)
#print pop[i]
#calculate the fitness of all individuals in the population
for i in range(0, pop_size):
fval[i][0] = i
fval[i][1] = fit.calculate_fitness(pop[i], cost)
#print fval[i][1]
initial_max_fit = min(fval[:, 1])
#print pop
print initial_max_fit
for k in range(0, generations):
new_pop = np.zeros(shape=(pop_size, cities))
#print 'hello'
#select parents from matting pool with tournament selection
for i in range(0, pop_size):
indx1 = random.randrange(0, pop_size)
indx2 = random.randrange(0, pop_size)
fit_val1 = fit.calculate_fitness(pop[indx1], cost)
