def GenerateLabel(data_path, name_list_path, time_message_path,
timeout_message_path):
with open(data_path, "r") as csvfile:
data = list(csv.reader(csvfile))
name_list = []
time_message = {}
timeout_message = {}
for line in data:
name_list.append(line[0])
time_dic = {}
timeout_dic = {}
for index in range(len(utils.method_list)):
if line[index +
1] == "timeout" or line[index +
1] == "failed" or line[index +
1] == "0.0":
time_dic[utils.method_list[index]] = 3600.0
timeout_dic[utils.method_list[index]] = True
else:
time_dic[utils.method_list[index]] = float(line[index + 1])
timeout_dic[utils.method_list[index]] = False
time_message[line[0]] = time_dic
timeout_message[line[0]] = timeout_dic
utils.WriteJson(name_list, name_list_path)
utils.WriteJson(time_message, time_message_path)
utils.WriteJson(timeout_message, timeout_message_path)
def GenerateLabelForOneMethod(data_path, name_list_path, time_dic_path,
timeout_dic_path, method):
method_list = utils.method_list
with open(data_path, "r") as csvfile:
data = list(csv.reader(csvfile))[1:]
name_list = []
time_dic = {}
timeout_dic = {}
for line in data:
name = line[0]
name_list.append(name)
index = method_list.index(method) + 1
if line[index] == "timeout" or line[index] == "failed" or line[
index] == "0.0" or line[index] == "0":
time_dic[name] = 3600.0
timeout_dic[name] = True
else:
time_dic[name] = float(line[index])
timeout_dic[name] = False
utils.WriteJson(name_list, name_list_path)
utils.WriteJson(time_dic, time_dic_path)
utils.WriteJson(timeout_dic, timeout_dic_path)
return name_list, time_dic, timeout_dic
def StatisticSamples(test_label_dic_path, statistic_name_dic_path):
test_label_dic = utils.ReadJson(test_label_dic_path)
statistic_name_dic = {}
for method in utils.method_list:
statistic_name_dic[method] = []
for name in test_label_dic.keys():
statistic_name_dic[test_label_dic[name]].append(name)
utils.WriteJson(statistic_name_dic, statistic_name_dic_path)
def GenerateLabel(data_path, name_list_path, label_dic_path):
with open(data_path, "r") as csvfile:
data = list(csv.reader(csvfile))
name_list = []
label_dic = {}
for line in data:
method = "None"
time = float(sys.maxsize)
for index in range(len(utils.method_list)):
if line[index +
1] != "timeout" and line[index + 1] != "failed" and line[
index + 1] != "0.0" and float(line[index + 1]) < time:
time = float(line[index + 1])
method = utils.method_list[index]
if method != "None":
name_list.append(line[0])
label_dic[line[0]] = method
else:
print(line[0])
utils.WriteJson(name_list, name_list_path)
utils.WriteJson(label_dic, label_dic_path)
def Predict(name_list_path, model_path, layer, method):
name_list = utils.ReadJson(name_list_path)
if layer == 0:
encoding_dic_dir = utils.encoding_dic_dir_0
elif layer == 1:
encoding_dic_dir = utils.encoding_dic_dir_1
elif layer == 2:
encoding_dic_dir = utils.encoding_dic_dir_2
vec_list = utils.GetVecListFromDic(encoding_dic_dir, name_list)
model = utils.Load_pkl(model_path)
predict_time_list = model.predict(vec_list)
time_predict = predict_time_list.tolist()
time_predict_path = utils.time_predict_path + "predict_" + method + "_" + str(layer) + ".json"
utils.WriteJson(time_predict, time_predict_path)
return time_predict
def GenerateEncodingDic(encoding_dir, encoding_dic_dir):
encoding_dic = {}
file_name_list = os.listdir(encoding_dir)
for file_name in file_name_list:
aig_name = file_name.split(".vector")[0]
file_path = os.path.join(encoding_dir, file_name)
assert (os.path.isfile(file_path))
vector = []
with open(file_path, encoding='utf-8') as fp:
line = fp.readlines()[0].split("[")[1].split("]")[0]
items = line.split(", ")
for item in items:
vector.append(int(item))
fp.close()
encoding_dic[aig_name] = vector
utils.WriteJson(encoding_dic, encoding_dic_dir)
def StatisticAvgEncodingTime(log_path, time_path):
dic = {}
with open(log_path, 'r') as file_obj:
data = file_obj.read()
file_obj.close()
lines = data.split("\n")
num = int(len(lines) / 2)
for i in range(num):
name_message = lines[2 * i]
time_message = lines[2 * i + 1]
name = name_message.split("../networks_aag/")[1].split(".aag")[0]
temp_time = time_message.split("user\t")[1].split("s")[0]
minute = float(temp_time.split("m")[0])
second = float(temp_time.split("m")[1])
time = minute * 60 + second
dic[name] = time
utils.WriteJson(dic, time_path)
return dic
def RandomForest(embedded_dir,
train_name_list,
test_name_list,
train_time_message,
test_time_message,
time_predict_path,
layer,
max_depth,
min_samples_split=2,
min_samples_leaf=1):
time_model_path = utils.time_model_path
train_vec_list = utils.GetVecList(embedded_dir, train_name_list)
test_vec_list = utils.GetVecList(embedded_dir, test_name_list)
time_predict_message = {}
for method in utils.method_list:
whole_time_model_path = time_model_path + method + "_model_" + layer + ".pkl"
train_time_list = GetTimeList(train_name_list, train_time_message,
method)
test_time_list = GetTimeList(test_name_list, test_time_message, method)
model = RandomForestRegressor(n_estimators=100,
criterion="mae",
max_depth=max_depth,
min_samples_split=min_samples_split,
min_samples_leaf=min_samples_leaf)
model.fit(train_vec_list, train_time_list)
utils.Save_pkl(model, whole_time_model_path)
predict_time_list = model.predict(test_vec_list)
time_predict_message[method] = predict_time_list.tolist()
print("Traing Score:%f" % model.score(train_vec_list, train_time_list))
print("Testing Score:%f" % model.score(test_vec_list, test_time_list))
# trees = model.estimators_
# dot_data = tree.export_graphviz(trees[0],
# out_file = None,
# feature_names = list(range(len(train_vec_list[0]))),
# class_names = train_time_list,
# filled = True,
# rounded = True
# )
# graph = pydotplus.graph_from_dot_data(dot_data)
# graph.write_pdf(utils.time_result_path + method + "_" + layer + ".pdf")
utils.WriteJson(time_predict_message, time_predict_path)
def RandomForest(embedded_dir, train_name_list, test_name_list, train_time_message, test_time_message,
time_predict_path, layer, max_depth, min_samples_split=2, min_samples_leaf=1):
time_model_path = utils.time_model_path
train_vec_list = utils.GetVecList(embedded_dir, train_name_list)
test_vec_list = utils.GetVecList(embedded_dir, test_name_list)
time_predict_message = {}
for method in utils.method_list:
whole_time_model_path = time_model_path + method + "_model_" + layer + ".pkl"
train_time_list = GetTimeList(train_name_list, train_time_message, method)
test_time_list = GetTimeList(test_name_list, test_time_message, method)
model = RandomForestRegressor(n_estimators=100, criterion="mae", max_depth=max_depth, min_samples_split=min_samples_split, min_samples_leaf=min_samples_leaf)
model.fit(train_vec_list, train_time_list)
utils.Save_pkl(model, whole_time_model_path)
predict_time_list = model.predict(test_vec_list)
time_predict_message[method] = predict_time_list.tolist()
print("Traing Score:%f" % model.score(train_vec_list, train_time_list))
print("Testing Score:%f" % model.score(test_vec_list, test_time_list))
utils.WriteJson(time_predict_message, time_predict_path)
def RandomForest(layer, embedded_dir, train_name_list, test_name_list, train_label_list, test_label_list, classify_predict_path,
model_path, importance_path, max_depth, max_leaf_nodes, min_samples_split=2, min_samples_leaf=1):
train_vec_list = utils.GetVecList(embedded_dir, train_name_list)
test_vec_list = utils.GetVecList(embedded_dir, test_name_list)
model = RandomForestClassifier(n_estimators=100, criterion="entropy", max_depth=None, min_samples_split=min_samples_split,
min_samples_leaf=min_samples_leaf, max_leaf_nodes=max_leaf_nodes, max_features=None)
classifier = model.fit(train_vec_list, train_label_list)
utils.Save_pkl(classifier, model_path)
predictions = classifier.predict_proba(test_vec_list)
predict_label_list = np.argsort(-predictions, axis=1)
importance = model.feature_importances_
utils.WriteJson(importance.tolist(), importance_path)
print("layer " + layer)
sum_acc = GetAcc(predict_label_list, test_label_list, utils.choose_top_method_number_1)
print(sum_acc)
sum_acc = GetAcc(predict_label_list, test_label_list, utils.choose_top_method_number_2)
print(sum_acc)
classify_predict = GeneratePredictResult(test_name_list, predict_label_list, classify_predict_path)
def GeneratePredictResult(test_name_list, predict_label_list, classify_predict_path):
classify_predict = {}
method_list = utils.method_list
for i in range(len(test_name_list)):
test_name = test_name_list[i]
temp_predict = []
for predict in predict_label_list[i]:
temp_predict.append(method_list[predict])
classify_predict[test_name] = temp_predict
utils.WriteJson(classify_predict, classify_predict_path)
# method_list = utils.method_list
# choose_top_method_number = utils.choose_top_method_number_1
# for i in range(choose_top_method_number):
# top_i_method_dic = {}
# for j in range(len(test_name_list)):
# test_name = test_name_list[j]
# top_i_method_dic[test_name] = classify_predict[test_name][i]
# statistic_dic = utils.Statistic([top_i_method_dic])
# print(statistic_dic)
return classify_predict
new_directory.append(reverse_item)
lettermark = False
firstmark = True
for i in range(len(reverse_item)):
if reverse_item[i] != "-":
if lettermark == False:
lettermark = True
if firstmark == False:
latex_format += "|"
else:
firstmark = False
latex_format += "\\mathtt{"
latex_format += reverse_item[i]
else:
if lettermark == True:
lettermark = False
latex_format += "}\\verb|"
latex_format += "-"
if lettermark == True:
latex_format += "}"
else:
latex_format += "|"
if index != len(directory) - 1:
latex_format += ", "
utils.WriteJson(new_directory, new_directory_path)
print(latex_format)
utils.WriteJson(latex_format, latex_format_path)
for method in utils.method_list:
save_path = utils.time_result_path + method + "_time_predict.pdf"
plt.figure()
plt.title(utils.NameMap(method), size=17)
# plt.xlabel('Sum Time (s)', size=25)
# plt.ylabel('# Solved Number', size=25)
for i in range(len(time_predict_label_list)):
label = time_predict_label_list[i]
time_predict = time_predict_list[i]
predict_time_list = time_predict[method]
predict_name_list = Sort(test_name_list, predict_time_list)
temp_name_sort_path = utils.time_basic_data_path + method + "_name_sort_" + str(
2 - i) + ".json"
utils.WriteJson(predict_name_list, temp_name_sort_path)
predict_solved_sum_time_list, predict_solved_number_list = GetFigData(
predict_name_list, method, test_time_message,
test_timeout_message)
if label == "0-depth Encoding":
color = "#9dc6e0"
color = "#a6a6a6"
elif label == "1-depth Encoding":
color = "#6996b3"
color = "#505050"
elif label == "2-depth Encoding":
# color = "#004c6d"
color = "#000000"
plt.plot(predict_solved_sum_time_list,
predict_solved_number_list,