def test_lightgbm():
from hyperactive import RandomSearchOptimizer
opt = RandomSearchOptimizer(search_config, 1)
opt.fit(X, y)
opt.predict(X)
opt.score(X, y)
def test_keras_losses():
from hyperactive import RandomSearchOptimizer
ml_losses = [
"mean_squared_error",
"mean_absolute_error",
"mean_absolute_percentage_error",
"mean_squared_logarithmic_error",
"squared_hinge",
"hinge",
# "categorical_hinge",
"logcosh",
"categorical_crossentropy",
# "sparse_categorical_crossentropy",
"binary_crossentropy",
"kullback_leibler_divergence",
"poisson",
"cosine_proximity",
]
for loss in ml_losses:
opt = RandomSearchOptimizer(search_config, 1, metric=loss)
assert opt._config_.metric == loss
opt.fit(X, y)
assert opt._config_.metric == loss
opt.predict(X)
assert opt._config_.metric == loss
opt.score(X, y)
assert opt._config_.metric == loss
def test_lightgbm_classification():
from hyperactive import RandomSearchOptimizer
ml_scores = [
"accuracy_score",
"balanced_accuracy_score",
"average_precision_score",
"brier_score_loss",
"f1_score",
"log_loss",
"precision_score",
"recall_score",
"jaccard_score",
"roc_auc_score",
]
for score in ml_scores:
opt = RandomSearchOptimizer(search_config, 1, metric=score)
assert opt._config_.metric == score
opt.fit(X, y)
assert opt._config_.metric == score
opt.predict(X)
assert opt._config_.metric == score
opt.score(X, y)
assert opt._config_.metric == score
def test_all_methods():
from hyperactive import RandomSearchOptimizer
Optimizer = RandomSearchOptimizer(search_config, n_iter=10, verbosity=0)
Optimizer.fit(X_train, y_train)
Optimizer.predict(X_test)
Optimizer.score(X_test, y_test)
Optimizer.export("test")
def test_lightgbm_verbosity():
from hyperactive import RandomSearchOptimizer
verbosity_list = [0, 1, 2]
for verbosity in verbosity_list:
opt = RandomSearchOptimizer(search_config, 1, verbosity=verbosity)
opt.fit(X, y)
opt.predict(X)
opt.score(X, y)
def test_lightgbm_random_state():
from hyperactive import RandomSearchOptimizer
random_state_list = [None, 0, 1, 2]
for random_state in random_state_list:
opt = RandomSearchOptimizer(search_config, 1, random_state=random_state)
opt.fit(X, y)
opt.predict(X)
opt.score(X, y)
def test_lightgbm_memory():
from hyperactive import RandomSearchOptimizer
memory_list = [False, True]
for memory in memory_list:
opt = RandomSearchOptimizer(search_config, 1, memory=memory)
opt.fit(X, y)
opt.predict(X)
opt.score(X, y)
def test_lightgbm_scatter_init():
from hyperactive import RandomSearchOptimizer
scatter_init_list = [False, 2, 3, 4]
for scatter_init in scatter_init_list:
opt = RandomSearchOptimizer(search_config, 1, scatter_init=scatter_init)
opt.fit(X, y)
opt.predict(X)
opt.score(X, y)
def test_lightgbm_n_iter():
from hyperactive import RandomSearchOptimizer
n_iter_list = [0, 1, 3, 10]
for n_iter in n_iter_list:
opt = RandomSearchOptimizer(search_config, n_iter)
opt.fit(X, y)
opt.predict(X)
opt.score(X, y)
def test_lightgbm_cv():
from hyperactive import RandomSearchOptimizer
cv_list = [0.1, 0.5, 0.9, 2, 4]
for cv in cv_list:
opt = RandomSearchOptimizer(search_config, 1, cv=cv)
opt.fit(X, y)
opt.predict(X)
opt.score(X, y)
def test_xgboost_warm_start():
from hyperactive import RandomSearchOptimizer
warm_start = {"sklearn.tree.DecisionTreeClassifier": {"max_depth": [1]}}
warm_start_list = [None, warm_start]
for warm_start in warm_start_list:
opt = RandomSearchOptimizer(search_config, 1, warm_start=warm_start)
opt.fit(X, y)
opt.predict(X)
opt.score(X, y)
def test_sklearn_cv():
from hyperactive import RandomSearchOptimizer
cv_list = [0.1, 0.5, 0.9, 2, 4]
for cv in cv_list:
opt = RandomSearchOptimizer(search_config, 1, cv=cv)
assert opt._config_.cv == cv
opt.fit(X, y)
assert opt._config_.cv == cv
opt.predict(X)
assert opt._config_.cv == cv
opt.score(X, y)
assert opt._config_.cv == cv
def test_sklearn_n_iter():
from hyperactive import RandomSearchOptimizer
n_iter_list = [0, 1, 3, 10]
for n_iter in n_iter_list:
opt = RandomSearchOptimizer(search_config, n_iter)
assert opt._config_.n_iter == n_iter
opt.fit(X, y)
assert opt._config_.n_iter == n_iter
opt.predict(X)
assert opt._config_.n_iter == n_iter
opt.score(X, y)
assert opt._config_.n_iter == n_iter
def test_sklearn_memory():
from hyperactive import RandomSearchOptimizer
memory_list = [False, True]
for memory in memory_list:
opt = RandomSearchOptimizer(search_config, 1, memory=memory)
assert opt._config_.memory == memory
opt.fit(X, y)
assert opt._config_.memory == memory
opt.predict(X)
assert opt._config_.memory == memory
opt.score(X, y)
assert opt._config_.memory == memory
def test_sklearn_random_state():
from hyperactive import RandomSearchOptimizer
random_state_list = [None, 0, 1, 2]
for random_state in random_state_list:
opt = RandomSearchOptimizer(search_config,
1,
random_state=random_state)
assert opt._config_.random_state == random_state
opt.fit(X, y)
assert opt._config_.random_state == random_state
opt.predict(X)
assert opt._config_.random_state == random_state
opt.score(X, y)
assert opt._config_.random_state == random_state
def test_sklearn_scatter_init():
from hyperactive import RandomSearchOptimizer
scatter_init_list = [False, 2, 3, 4]
for scatter_init in scatter_init_list:
opt = RandomSearchOptimizer(search_config,
1,
scatter_init=scatter_init)
assert opt._config_.scatter_init == scatter_init
opt.fit(X, y)
assert opt._config_.scatter_init == scatter_init
opt.predict(X)
assert opt._config_.scatter_init == scatter_init
opt.score(X, y)
assert opt._config_.scatter_init == scatter_init
def test_lightgbm_warm_start():
from hyperactive import RandomSearchOptimizer
warm_start = {
"lightgbm.LGBMClassifier": {
"boosting_type": ["gbdt"],
"num_leaves": [5],
"learning_rate": [0.05],
}
}
warm_start_list = [None, warm_start]
for warm_start in warm_start_list:
opt = RandomSearchOptimizer(search_config, 1, warm_start=warm_start)
opt.fit(X, y)
opt.predict(X)
opt.score(X, y)
def test_catboost_warm_start():
from hyperactive import RandomSearchOptimizer
warm_start = {
"catboost.CatBoostClassifier": {
"iterations": [3],
"learning_rate": [1],
"depth": [3],
"verbose": [0],
}
}
warm_start_list = [None, warm_start]
for warm_start in warm_start_list:
opt = RandomSearchOptimizer(search_config, 1, warm_start=warm_start)
opt.fit(X, y)
opt.predict(X)
opt.score(X, y)
def test_keras_scores():
from hyperactive import RandomSearchOptimizer
ml_scores = [
"accuracy",
"binary_accuracy",
"categorical_accuracy",
"sparse_categorical_accuracy",
"top_k_categorical_accuracy",
"sparse_top_k_categorical_accuracy",
]
for score in ml_scores:
opt = RandomSearchOptimizer(search_config, 1, metric=score)
assert opt._config_.metric == score
opt.fit(X, y)
assert opt._config_.metric == score
opt.predict(X)
assert opt._config_.metric == score
opt.score(X, y)
assert opt._config_.metric == score
def test_lightgbm_regression():
from hyperactive import RandomSearchOptimizer
ml_losses = [
"explained_variance_score",
"max_error",
"mean_absolute_error",
"mean_squared_error",
"mean_squared_log_error",
"median_absolute_error",
"r2_score",
]
for loss in ml_losses:
opt = RandomSearchOptimizer(search_config, 1, metric=loss)
assert opt._config_.metric == loss
opt.fit(X, y)
assert opt._config_.metric == loss
opt.predict(X)
assert opt._config_.metric == loss
opt.score(X, y)
assert opt._config_.metric == loss
cancer_data = load_breast_cancer()
X = cancer_data.data
y = cancer_data.target
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.20)
# this defines the model and hyperparameter search space
search_config = {
"lightgbm.LGBMClassifier": {
"boosting_type": ["gbdt"],
"num_leaves": range(2, 20),
"learning_rate": np.arange(0.01, 0.1, 0.01),
"feature_fraction": np.arange(0.1, 0.95, 0.1),
"bagging_fraction": np.arange(0.1, 0.95, 0.1),
"bagging_freq": range(2, 10, 1),
}
}
opt = RandomSearchOptimizer(search_config, n_iter=10, n_jobs=4, cv=3)
# search best hyperparameter for given data
opt.fit(X, y)
# predict from test data
prediction = opt.predict(X_test)
# calculate score
score = opt.score(X_test, y_test)
print("\ntest score of best model:", score)
"kernel_size": [3],
"activation": ["relu"],
},
"keras.layers.MaxPooling2D.4": {
"pool_size": [(2, 2)]
},
"keras.layers.Flatten.5": {},
"keras.layers.Dense.6": {
"units": range(10, 200, 10),
"activation": ["softmax"]
},
"keras.layers.Dropout.7": {
"rate": np.arange(0.2, 0.8, 0.1)
},
"keras.layers.Dense.8": {
"units": [10],
"activation": ["softmax"]
},
}
Optimizer = RandomSearchOptimizer(search_config, n_iter=10, metric="accuracy")
# search best hyperparameter for given data
Optimizer.fit(X_train, y_train)
# predict from test data
prediction = Optimizer.predict(X_test)
# calculate score
score = Optimizer.score(X_test, y_test)