def execute_example(distance_matrix, algorithm, propose_type, path1, path2,
history):
dist_1 = calculate_distance(distance_matrix, path1)
dist_2 = calculate_distance(distance_matrix, path2)
starting_dist = dist_1 + dist_2
path1, path2, history = algorithm(distance_matrix, path1, path2,
propose_type, history)
dist_1 = calculate_distance(distance_matrix, path1)
dist_2 = calculate_distance(distance_matrix, path2)
return starting_dist, (dist_1 + dist_2)
def execute_example_random(distance_matrix, algorithm, propose_type,
starting_history, picked_nodes):
path1 = starting_history[-1][0]
path2 = starting_history[-1][1]
dist_1 = calculate_distance(distance_matrix, path1)
dist_2 = calculate_distance(distance_matrix, path2)
starting_dist = dist_1 + dist_2
path1, path2, history = algorithm(distance_matrix, path1, path2,
propose_type, starting_history)
dist_1 = calculate_distance(distance_matrix, path1)
dist_2 = calculate_distance(distance_matrix, path2)
return starting_dist, (dist_1 + dist_2), picked_nodes
def k_regret_connector(algos, distance_matrix, start_with=None, k=1):
cycles = sample(list(range(len(distance_matrix))), len(algos))
if start_with is not None:
cycles = start_with
picked_nodes = deepcopy(cycles)
cycles = [[c] for c in cycles]
visited = []
for c in cycles:
visited += c
history = [deepcopy(cycles)]
enough = []
while sum([len(c) for c in cycles]) < len(distance_matrix):
min_regret = int(10e20)
best_algo = -1
for i, algo in enumerate(algos):
if i in enough:
continue
best_cycle, best_node = algo(distance_matrix, visited, cycles[i])
best_cost = calculate_distance(distance_matrix,
best_cycle) - calculate_distance(
distance_matrix, cycles[i])
regret = 0
sub_visited = visited + [best_node]
for ik in range(k - 1):
sub_cycle, sub_node = algo(distance_matrix, sub_visited,
cycles[i])
sub_cost = calculate_distance(distance_matrix,
sub_cycle) - calculate_distance(
distance_matrix, cycles[i])
regret += sub_cost - best_cost
sub_visited.append(sub_node)
if regret < min_regret:
min_regret = regret
best_algo = i
res_cycle, res_node = algos[best_algo](distance_matrix, visited,
cycles[best_algo])
cycles[best_algo] = deepcopy(res_cycle)
visited.append(res_node)
history.append(deepcopy(cycles))
if len(cycles[best_algo]) >= len(distance_matrix) // len(cycles) \
and len(enough) + 1 != len(cycles):
enough.append(best_algo)
return history, picked_nodes
res = k_regret(distance_matrix)[-1]
start = time()
checkpoint = time()
done = False
while True:
if not (k_regret_only and done):
res = steepest(distance_matrix, res[0], res[1])[-1]
done = True
checkpoint = time()
if checkpoint-start > time_allowed:
break
perturbateBig_(res[0], res[1])
res = k_regret(distance_matrix, start_with=res)[-1]
return res
if __name__ == '__main__':
data_set = 'kroB'
overview, coordinates = read_file('./data/' + data_set + '200.tsp')
distance_matrix = count_dist(coordinates)
# # MSLS:
# msls_results = []
# for i in range(10):
c1, c2 = ils1(distance_matrix)
dist = calculate_distance(distance_matrix, c1) + \
calculate_distance(distance_matrix,c2)
print(dist)
animate([c1, c2], coordinates, isCycle=[True, True])