def local_search_accept_error(nb_step=10, version=2, p=.1):
cur_sol = (ls.first_solution(graph, terminals))
act_gain = ls.gain(cur_sol)
l_act = [act_gain]
l_new = [act_gain]
for i in range(nb_step):
print(i)
new_sol = ls.neighbors_of_solution(graph, cur_sol, terminals)
new_gain = ls.gain(new_sol)
l_new.append(new_gain)
r = random.random()
if new_gain < act_gain or r < p:
act_gain = new_gain
cur_sol = new_sol
l_act.append(act_gain)
return l_act, l_new
def variation_crossover(terminals, population, lda):
for i in range(lda):
to_merge = random.sample(population, 2)
population.append(fusion(to_merge[0][1], to_merge[1][1], terminals))
population[-1] = (ls.gain(population[-1]), population[-1],
to_merge[0][2] + to_merge[1][2] + 1)
return population
def random_solution(graph, terminals, n):
sols = []
first_sol = ls.first_solution(graph, terminals)
for i in range(n):
n_graph = ls.neighbors_of_solution(graph, first_sol, terminals, 10)
poids = ls.gain(n_graph)
sols.append((poids, n_graph, 0))
return sols
def variation_mutation(terminals, population, lda):
for i in range(lda):
to_mutate = random.sample(population, 1)
population.append(
ls.neighbors_of_solution(graph, to_mutate[0][1], terminals, 2, 1))
population[-1] = (ls.gain(population[-1]), population[-1],
to_mutate[0][2] + 1)
return population
def local_search_only_better(nb_step=10, version=2):
cur_sol = (ls.first_solution(graph, terminals))
act_gain = ls.gain(cur_sol)
l_act = [act_gain]
l_new = [act_gain]
for i in range(nb_step):
print(i)
new_sol = ls.neighbors_of_solution(graph, cur_sol, terminals, version,
5)
new_gain = ls.gain(new_sol)
l_new.append(new_gain)
if new_gain < act_gain:
act_gain = new_gain
cur_sol = new_sol
l_act.append(act_gain)
print(ls.final_value(cur_sol))
print(ls.gain(cur_sol))
return l_act, l_new
def simulated_anhilling(nb_step=10, heat_strategy=heat_strategy_linear):
cur_sol = (ls.first_solution(graph, terminals))
act_gain = ls.gain(cur_sol)
l_act = [act_gain]
l_new = [act_gain]
l_seuils = [1]
for nb_step_act in range(nb_step):
new_sol = ls.neighbors_of_solution(graph, cur_sol, terminals)
new_gain = ls.gain(new_sol)
l_new.append(new_gain)
r = random.random()
r_seuil = heat_strategy(nb_step_act + 1, act_gain, new_gain)
l_seuils.append(r_seuil)
if r < r_seuil:
act_gain = new_gain
cur_sol = new_sol
l_act.append(act_gain)
return l_act, l_new, l_seuils