def run_anneal(init_function,move_operator,objective_function,max_iterations,start_temp,alpha):
if start_temp is None or alpha is None:
usage();
print "missing --cooling start_temp:alpha for annealing"
sys.exit(1)
from sa import anneal
iterations,score,best=anneal(init_function,move_operator,objective_function,max_iterations,start_temp,alpha)
return iterations,score,best
def run_anneal(init_function,move_operator,objective_function,max_iterations,start_temp,alpha):
if start_temp is None or alpha is None:
usage();
print "missing --cooling start_temp:alpha for annealing"
sys.exit(1)
from sa import anneal
iterations,score,best=anneal(init_function,move_operator,objective_function,max_iterations,start_temp,alpha)
return iterations,score,best
def test_codec():
#run a simple test: code challenge
N=300
data=loadtxt('example.txt')
J=coo_matrix((data[:,2],(data[:,0],data[:,1])),shape=(N,N),dtype='int32').toarray()
J=(J+J.T)/2.
cc=CC(J)
Emin,Config=anneal(cc,tempscales=linspace(10,0.6,51),nms=4000,nrun=30)
assert_(Emin==-3858 and cc.get_cost(Config)==Emin)
def run_sa(iterations, dist, start, alpha, draw=True, every=100):
init_function = lambda: init_random_tour(len(dist[0]))
objective_function = lambda tour: -tour_length(dist, tour)
move_operator = reversed_sections
iterations,score,best = sa.anneal(init_function, move_operator, \
objective_function, iterations, \
start, alpha, draw, every)
print iterations, score, best
return best
def run_sa(iterations, dist, start, alpha, draw = True, every = 100):
init_function = lambda: init_random_tour(len(dist[0]))
objective_function = lambda tour: -tour_length(dist, tour)
move_operator = reversed_sections
iterations,score,best = sa.anneal(init_function, move_operator, \
objective_function, iterations, \
start, alpha, draw, every)
print iterations, score, best
return best
def solve_anneal(aff):
objective_function=lambda tour: tour_sum_affinities(aff,tour)
init_function=lambda: tsp.init_random_tour(sqrt(len(aff)))
move_operator=tsp.reversed_sections
#suggested parameters for simulated annealing:
start_temp = 10.0
max_iterations = 5000
alpha = .001**(1/max_iterations)
from sa import anneal
iterations,score,best=anneal(init_function,move_operator,objective_function,max_iterations,start_temp,alpha)
return iterations,score,best
def test_simple_anneal():
'''
check we perform the same as a hillclimb
when score keeps improving
'''
def move_operator(i):
yield i+1
def init_function():
return 1
num_evaluations,best_score,best=sa.anneal(init_function,move_operator,objective_function,max_evaluations,1,0.9)
assert num_evaluations == max_evaluations
assert best == max_evaluations
assert best_score == max_evaluations
print("Working with {} engines\n".format(len(engines)))
print("Initial state")
sa.print_state(state, current=True)
eg_thrust = 50
eg_turn = 10
eg_heat = 10
eg_capacity = 20
eg_mass = 20
print(f"Trial th{eg_thrust},tn{eg_turn},h{eg_heat},c{eg_capacity},m{eg_mass}:")
sa.print_state(
sa.anneal(copy.deepcopy(state),
iterations,
engines,
eg_thrust,
eg_turn,
eg_heat,
eg_capacity,
eg_mass,
verbose=False))
eg_thrust = 30
eg_turn = 15
eg_heat = 20
eg_capacity = 5
eg_mass = 20
print(f"Trial th{eg_thrust},tn{eg_turn},h{eg_heat},c{eg_capacity},m{eg_mass}:")
sa.print_state(
sa.anneal(copy.deepcopy(state),
iterations,
engines,
