def trainabGA(model_name,dataset_name,iter_num):
# init T,R,SP
T = getSGGRes(model_name,dataset_name,"train",tasktype)
R = getGT(model_name,dataset_name,"train",tasktype)
SP,N = getPfromR(R,dataset_name)
print("T info: ",len(T),len(T[0]),T[0][0])
print("R info: ",len(R),len(R[0]),R[0][0])
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
# start iteration
low = [0.0,0.0]
high = [1e6,1e6]
for i in range(iter_num):
#ga = GA(func=fitness, n_dim=2, size_pop=4, max_iter=10, \
# lb=[0.0, 0.0], ub=[1.0, 1.0], precision=1e-7)
#ga = GA(func=fitness, n_dim=2, size_pop=6, max_iter=50, \
# lb=low, ub=high, precision=1e-7)
#ga = GA(func=fitness, n_dim=2, size_pop=50, max_iter=50, \
# lb=low, ub=high, precision=1e-7)
ga = GA(func=fitnessNDCG, n_dim=2, size_pop=100, max_iter=100, \
lb=low, ub=high, precision=1e-7)
ga.register(operator_name='selection', operator=selection_tournament, tourn_size=3).\
register(operator_name='ranking', operator=ranking.ranking). \
register(operator_name='crossover', operator=crossover.crossover_2point). \
register(operator_name='mutation', operator=mutation.mutation)
ga.to(device=device) #GPU
start_time = time.time()
best_x, best_y = ga.run()# best_x={alpha,beta},best_y=Q-R
print("run time: ",time.time() - start_time)
alpha, beta = best_x
print('iteration ',i,': best alpha and beta: ', alpha,beta)
print('best_y:', best_y)
print("sum_ndcg: ",-best_y-0.5*(alpha**2+beta**2))
if best_y==0:
print('best {alpha,beta}:',best_x,'best_y',best_y)
end_cnt = fitnessNDCG(best_x)
print("detect if zeros: ",end_cnt)
if end_cnt==0:
break
else:
#R = ReSet(alpha,beta)
R = addsmallNDCG(alpha,beta,R)
SP,_ = getPfromR(R,dataset_name)
cnt_now=fitness(best_x)
print(i," iter :", cnt_now)
return best_x
def config_ga():
lb = [1] * (station_count + truck_count - 1)
ub = [21] * (station_count + truck_count - 1)
iteration = int(cfg.read_config("iteration_time"))
ga = GA(func=fitness_fun,
size_pop=100,
n_dim=station_count + truck_count - 1,
max_iter=iteration,
prob_mut=0.01,
lb=lb,
ub=ub,
precision=[1] * (station_count + truck_count - 1))
ga.Chrom = generate_new_pop()
if cfg.read_config("crossover_type") is not None:
ga.register(operator_name='crossover', operator=customized_crossover). \
register(operator_name='mutation', operator=customized_mutation)
# TODO: 排序的策略配置（暂时不考虑）
return ga
def config_ga():
constraint_eq = [
lambda x: int(cfg.read_config('bike_total_count')) - sum(x)
]
lb = [0] * station_count
ub = hub.hub
iteration = int(cfg.read_config("iteration_time"))
ga = GA(func=fitness_fun,
size_pop=300,
n_dim=station_count,
max_iter=iteration,
prob_mut=0.01,
lb=lb,
ub=ub,
constraint_eq=constraint_eq,
precision=[1] * 20)
if cfg.read_config("crossover_type") is not None:
ga.register(operator_name='crossover',
operator=crossover.crossover_1point)
# TODO: 变异和排序的策略配置（暂时不考虑）
return ga
# %% step2: import package and build ga, as usual.
import numpy as np
from sko.GA import GA, GA_TSP
demo_func = lambda x: x[0]**2 + (x[1] - 0.05)**2 + (x[2] - 0.5)**2
ga = GA(func=demo_func,
n_dim=3,
size_pop=100,
max_iter=500,
lb=[-1, -10, -5],
ub=[2, 10, 2],
precision=[1e-7, 1e-7, 1])
# %% step3: register your own operator
ga.register(operator_name='selection',
operator=selection_tournament,
tourn_size=3)
# %% Or import the operators scikit-opt already defined.
from sko.operators import ranking, selection, crossover, mutation
ga.register(operator_name='ranking', operator=ranking.ranking). \
register(operator_name='crossover', operator=crossover.crossover_2point). \
register(operator_name='mutation', operator=mutation.mutation)
# %% Run ga
best_x, best_y = ga.run()
print('best_x:', best_x, '\n', 'best_y:', best_y)
# %% For advanced users
class MyGA(GA):
def selection(self, tourn_size=3):