def test_SelectPercentile(self):
scenario, sampler, rules = get_configuration_SelectPercentile()
def evaluate(config, bestconfig):
try:
for name, params in config:
if name == "SelectPercentile":
pipeline = Pipeline(
steps=[(name,
feature_selection.SelectPercentile(
**params)),
("logistic_regression",
linear_model.LogisticRegression())])
pipeline = fit_model(pipeline, X_train, y_train)
return pipeline.score(X_test, y_test)
raise Exception("Classifier not found")
except Exception:
return 0
except TimedOutExc:
return 0
searcher = Search(scenario, sampler, rules, evaluate)
searcher.run(
nb_simulation=10,
generate_image_path="out/data_preprocessing/SelectPercentile")
def test_DictionaryLearning(self):
scenario, sampler, rules = get_configuration_DictionaryLearning()
def evaluate(config, bestconfig):
try:
for name, params in config:
if name == "DictionaryLearning":
pipeline = Pipeline(
steps=[(name,
decomposition.DictionaryLearning(
**params)),
("logistic_regression",
linear_model.LogisticRegression())])
pipeline = fit_model(pipeline, X_train, y_train)
return pipeline.score(X_test, y_test)
raise Exception("Classifier not found")
except ValueError:
return 0
except TimedOutExc:
return 0
searcher = Search(scenario, sampler, rules, evaluate)
searcher.run(
nb_simulation=10,
generate_image_path="out/data_preprocessing/DictionaryLearning")
def test_FunctionTransformer(self):
scenario, sampler, rules = get_configuration_FunctionTransformer()
def evaluate(config, bestconfig):
try:
for name, params in config:
if name == "FunctionTransformer":
pipeline = Pipeline(
steps=[(name,
preprocessing.FunctionTransformer(
**params)),
("logistic_regression",
linear_model.LogisticRegression())])
pipeline = fit_model(pipeline, X_train, y_train)
return pipeline.score(X_test, y_test)
raise Exception("Classifier not found")
except ValueError:
return 0
except pynisher.TimeoutException:
return 0
searcher = Search(scenario, sampler, rules, evaluate)
searcher.run(
nb_simulation=10,
generate_image_path="out/data_preprocessing/FunctionTransformer")
return r2
if __name__ == "__main__":
# iris = datasets.load_iris()
# X_train, X_test, y_train, y_test = train_test_split(
# iris.data, iris.target, test_size=0.33, random_state=42)
# environment = Environment(svm_from_cfg,
# config_space=configuration_space.cs,
# mem_in_mb=2048,
# cpu_time_in_s=30,
# seed=42)
environment = Environment(run_pipeline,
config_space=configuration_space.cs,
mem_in_mb=8000,
cpu_time_in_s=300000000,
seed=42)
mosaic = Search(environment=environment,
bandit_policy={
"policy_name": "uct",
"c_ucb": 1.1
},
coef_progressive_widening=0.6,
verbose=True)
best_config, best_score = mosaic.run(nb_simulation=10000)
print("Best config: ", best_config, "best score", best_score)
# For deactivated parameters, the configuration stores None-values.
# This is not accepted by the SVM, so we remove them.
cfg = {k: cfg[k] for k in cfg if cfg[k]}
# We translate boolean values:
cfg["shrinking"] = True if cfg["shrinking"] == "true" else False
# And for gamma, we set it to a fixed value or to "auto" (if used)
if "gamma" in cfg:
cfg["gamma"] = cfg["gamma_value"] if cfg["gamma"] == "value" else "auto"
cfg.pop("gamma_value", None) # Remove "gamma_value"
clf = svm.SVC(**cfg, random_state=42)
scores = cross_val_score(clf, X_train, y_train, cv=5)
return np.mean(scores) # Minimize!
environment = Environment(svm_from_cfg,
config_space=cs,
mem_in_mb=2048,
cpu_time_in_s=30,
seed=42)
mosaic = Search(environment=environment,
policy_arg={
"c_ucb": 1.1,
"coef_progressive_widening": 0.6
},
verbose=True)
best_config, best_score = mosaic.run(nb_simulation=100)
print("Best config: ", best_config, "best score", best_score)