def testingILS(original_matrix, original_samples):
bestResult = np.inf
bestTime = 0
bestGroup = []
startGroup = []
regretGroups = []
random_groups = []
random_groups = greedMethod(20, original_matrix, original_samples)
regretGroups = steepest_cashe_and_list(random_groups, original_matrix)
for i in range(1000):
print("START", i)
start = time.time()
random_groups = small_perturbate(regretGroups, 10)
regretGroups = steepest_cashe_and_list(random_groups, original_matrix)
elapsedTime = time.time() - start
print(str(elapsedTime) + " sec")
result, _, _ = count_result(regretGroups, original_matrix)
print(str(result) + " result")
if bestResult > result:
bestResult = result
bestTime = elapsedTime
bestGroup = regretGroups.copy()
startGroup = random_groups.copy()
print("BEST")
print(bestTime)
print(bestResult)
print("BEST")
return bestGroup, startGroup
def compare_pairs_with_the_best(original_matrix, original_samples):
bestResult = np.inf
bestGroup = []
results = []
result_groups = []
similarity_with_best = []
similarity_with_others = []
for i in range(500):
print("START", i)
regretGroups = []
random_groups = []
random_groups = greedMethod(20, original_matrix, original_samples)
#regretGroups = greedy(random_groups.copy(), original_matrix)
regretGroups = steepest_cashe_and_list(random_groups, original_matrix)
result, _, _ = count_result(regretGroups, original_matrix)
print(str(result) + " result")
results.append(result)
result_groups.append(regretGroups)
if bestResult > result:
bestResult = result
bestGroup = regretGroups.copy()
best_id = results.index(bestResult)
result_groups.pop(best_id)
results.pop(best_id)
for group in result_groups:
similarity_with_best.append(compare_pairs2(group, bestGroup))
for i in range(len(result_groups)):
for j in range(i + 1, len(result_groups)):
result += compare_pairs2(result_groups[i], result_groups[j])
result = result / (len(result_groups) - 1)
similarity_with_others.append(result)
print(similarity_with_best, similarity_with_others, results)
with_best_perason = scipy.stats.pearsonr(results, similarity_with_best)
print(with_best_perason)
with_others_pearson = scipy.stats.pearsonr(results, similarity_with_others)
print(with_others_pearson)
plt.plot(results, similarity_with_best, 'ro')
plt.title("Podobieństwo względem najlepszego optimum lokalnego")
plt.xlabel("Wartość funkcji celu")
plt.ylabel("Podobieństwo")
plt.savefig('similarity_with_best.png')
plt.show()
plt.title("Średnie podobieństwo względem pozostałych optimów lokalnych")
plt.xlabel("Wartość funkcji celu")
plt.ylabel("Podobieństwo")
plt.plot(results, similarity_with_others, 'bo')
plt.savefig('similarity_with_others.png')
plt.show()
def testingMSLS(original_matrix, original_samples):
best_times = []
best_groups = []
best_results = []
for j in range(10):
bestResult = np.inf
bestTime = 0
bestGroup = []
startGroup = []
regretGroups = []
random_groups = []
for i in range(100):
print("START", j, i)
random_groups = greedMethod(20, original_matrix, original_samples)
start = time.time()
regretGroups = steepest_cashe_and_list(random_groups,
original_matrix)
elapsedTime = time.time() - start
print(str(elapsedTime) + " sec")
result, _, _ = count_result(regretGroups, original_matrix)
print(str(result) + " result")
if bestResult > result:
bestResult = result
bestTime = elapsedTime
bestGroup = regretGroups.copy()
startGroup = random_groups.copy()
print("BEST")
print(bestTime)
print(bestResult)
print("BEST")
best_times.append(bestTime)
best_groups.append(bestGroup.copy())
best_results.append(bestResult)
print("***RESULTS***")
print('MIN: ', np.min(best_results))
print('MAX: ', np.max(best_results))
print('MEAN: ', np.average(best_results))
print('STD: ', np.std(best_results))
print("***TIME***")
print('MIN: ', np.min(best_times))
print('MAX: ', np.max(best_times))
print('MEAN: ', np.average(best_times))
print('STD: ', np.std(best_times))
return bestGroup, startGroup
if __name__ == '__main__':
no_groups = 20
fileName = "data/objects20_06.data"
# fileName = "data/objects.data"
# fileName = "data/test.data"
# I read data from file
samples = loadInstance(fileName)
# II prepare coordinates data
x_samples = [pair[0] for pair in samples]
y_samples = [pair[1] for pair in samples]
# III calculate matrix with euclidean distances
matrix = prepareMatrix(samples)
# IV Groups from first project - Greedy
greed_groups = greedMethod(no_groups, matrix, samples)
print(count_result(greed_groups, matrix))
#drawRegret(x_samples, y_samples, greed_groups)
# V Start local search - steepest
# steepest_groups = steepest(greed_groups, matrix)
# drawRegret(x_samples, y_samples, steepest_groups)
# print(count_result(steepest_groups, matrix))
# TESTING MSLS
best, random = testingILS(matrix, samples)
drawRegret(x_samples, y_samples, random)
drawRegret(x_samples, y_samples, best)