def test_stratified_binary_classification():
X = cp.array([[0.37487513, -2.3031888, 1.662633, 0.7671007],
[-0.49796826, -1.0621182, -0.32518214, -0.20583323],
[-1.0104885, -2.4997945, 2.8952584, 1.4712684],
[2.008748, -2.4520662, 0.5557737, 0.07749569],
[0.97350526, -0.3403474, -0.58081895, -0.23199573]])
# Needs to fail when we have just 1 occurence of a label
y = cp.array([0, 0, 0, 0, 1])
with pytest.raises(ValueError):
train_test_split(X, y, train_size=0.75, stratify=y, shuffle=True)
y = cp.array([0, 0, 0, 1, 1])
X_train, X_test, y_train, y_test = train_test_split(X,
y,
train_size=0.75,
stratify=y,
random_state=15)
_, y_counts = cp.unique(y, return_counts=True)
_, train_counts = cp.unique(y_train, return_counts=True)
_, test_counts = cp.unique(y_test, return_counts=True)
# Ensure we have preserve the number of labels
cp.testing.assert_array_equal(train_counts + test_counts, y_counts)
def train_test_split(X,
y=None,
shuffle=True,
random_state=None,
stratify=None,
**kwargs):
if y is not None and str(y.dtype) == 'object':
return _train_test_split_with_object(X,
y,
shuffle=shuffle,
random_state=random_state,
stratify=stratify,
**kwargs)
try:
return cm_sel.train_test_split(X,
y,
shuffle=shuffle,
random_state=random_state,
stratify=stratify,
**kwargs)
except Exception as e:
if stratify is not None and str(e).find('cudaErrorInvalidValue') >= 0:
logger.warning('train_test_split failed, retry without stratify')
return cm_sel.train_test_split(X,
y,
shuffle=shuffle,
random_state=random_state,
**kwargs)
else:
raise
def test_random_state(seed_type):
for i in range(10):
seed_n = np.random.randint(0, int(1e9))
if seed_type == 'int':
seed = seed_n
if seed_type == 'cupy':
seed = cp.random.RandomState(seed=seed_n)
if seed_type == 'numpy':
seed = np.random.RandomState(seed=seed_n)
X = cudf.DataFrame({"x": range(100)})
y = cudf.Series(([0] * (100 // 2)) + ([1] * (100 // 2)))
X_train, X_test, y_train, y_test = train_test_split(X,
y,
random_state=seed)
if seed_type == 'cupy':
seed = cp.random.RandomState(seed=seed_n)
if seed_type == 'numpy':
seed = np.random.RandomState(seed=seed_n)
X_train2, X_test2, y_train2, y_test2 = \
train_test_split(X, y, random_state=seed)
assert X_train.equals(X_train2)
assert X_test.equals(X_test2)
assert y_train.equals(y_train2)
assert y_test.equals(y_test2)
def test_split_dataframe_array(y_type):
X = cudf.DataFrame({"x": range(100)})
y = cudf.Series(([0] * (100 // 2)) + ([1] * (100 // 2)))
if y_type == "cupy":
X_train, X_test, y_train, y_test = train_test_split(X, y.values)
assert isinstance(X_train, cudf.DataFrame)
assert isinstance(X_test, cudf.DataFrame)
assert isinstance(y_train, cp.ndarray)
assert isinstance(y_test, cp.ndarray)
elif y_type == "cudf":
X_train, X_test, y_train, y_test = train_test_split(X, y)
assert isinstance(X_train, cudf.DataFrame)
assert isinstance(X_test, cudf.DataFrame)
assert isinstance(y_train, cudf.Series)
assert isinstance(y_test, cudf.Series)
def make_classification_dataset(datatype, nrows, ncols, nclasses):
n_real_features = min(ncols, int(max(nclasses * 2, math.ceil(ncols / 10))))
n_clusters_per_class = min(2, max(1, int(2**n_real_features / nclasses)))
n_redundant = min(ncols - n_real_features, max(2, math.ceil(ncols / 20)))
try:
X, y = data.make_classification(
dtype=datatype,
n_samples=nrows + 1000,
n_features=ncols,
random_state=SEED,
class_sep=1.0,
n_informative=n_real_features,
n_clusters_per_class=n_clusters_per_class,
n_redundant=n_redundant,
n_classes=nclasses)
r = dsel.train_test_split(X, y, random_state=SEED, train_size=nrows)
if len(cp.unique(r[2])) < nclasses:
raise ValueError("Training data does not have all classes.")
return r
except ValueError:
pytest.skip(
"Skipping the test for invalid combination of ncols/nclasses")
def train_and_eval(X_param,
y_param,
penalty='l2',
C=1.0,
l1_ratio=None,
fit_intercept=True):
"""
Splits the given data into train and test split to train and evaluate the model
for the params parameters.
Params
______
X_param: DataFrame.
The data to use for training and testing.
y_param: Series.
The label for training
penalty, C, l1_ratio, fit_intercept: The parameter values for Logistic Regression.
Returns
score: log loss of the fitted model
"""
X_train, X_valid, y_train, y_valid = train_test_split(X_param,
y_param,
random_state=42)
classifier = LogisticRegression(penalty=penalty,
C=C,
l1_ratio=l1_ratio,
fit_intercept=fit_intercept)
classifier.fit(X_train, y_train)
y_pred = classifier.predict(X_valid)
score = log_loss(y_valid, y_pred)
return score
def test_split_array_single_argument(type, test_size, train_size, shuffle):
X = np.zeros((100, 10)) + np.arange(100).reshape(100, 1)
if type == 'cupy':
X = cp.asarray(X)
if type == 'numba':
X = cuda.to_device(X)
X_train, X_test = train_test_split(X,
train_size=train_size,
test_size=test_size,
shuffle=shuffle,
random_state=0)
if type == 'cupy':
assert isinstance(X_train, cp.ndarray)
assert isinstance(X_test, cp.ndarray)
if type in ['numba', 'rmm']:
assert cuda.devicearray.is_cuda_ndarray(X_train)
assert cuda.devicearray.is_cuda_ndarray(X_test)
if train_size is not None:
assert X_train.shape[0] == (int)(X.shape[0] * train_size)
if test_size is not None:
assert X_test.shape[0] == (int)(X.shape[0] * test_size)
if shuffle is None:
assert X_train == X[0:train_size]
assert X_test == X[-1 * test_size:]
X_rec = cp.sort(cp.concatenate(X_train, X_test))
assert X_rec == X
def test_pipeline():
X, y = make_classification(random_state=0)
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)
pipe = Pipeline(steps=[('scaler', StandardScaler()), ('svc', SVC())])
pipe.fit(X_train, y_train)
score = pipe.score(X_test, y_test)
assert score > 0.8
def test_split_column():
data = cudf.DataFrame(
{
"x": range(100),
"y": ([0] * (100 // 2)) + ([1] * (100 // 2)),
}
)
train_size = 0.8
X_train, X_test, y_train, y_test = train_test_split(
data, "y", train_size=train_size
)
assert (
len(X_train) == len(y_train) == pytest.approx(train_size * len(data))
)
assert (
len(X_test)
== len(y_test)
== pytest.approx((1 - train_size) * len(data))
)
X_reconstructed = cudf.concat([X_train, X_test]).sort_values(
by=["x"]
)
y_reconstructed = y_train.append(y_test).sort_values()
assert all(
data == X_reconstructed.assign(
y=y_reconstructed).reset_index(drop=True)
)
def test_sklearn_search():
"""Test ensures scoring function works with sklearn machinery
"""
import numpy as np
from cuml import Ridge as cumlRidge
import cudf
from sklearn import datasets
from sklearn.model_selection import train_test_split, GridSearchCV
diabetes = datasets.load_diabetes()
X_train, X_test, y_train, y_test = train_test_split(diabetes.data,
diabetes.target,
test_size=0.2,
shuffle=False,
random_state=1)
alpha = np.array([1.0])
fit_intercept = True
normalize = False
params = {'alpha': np.logspace(-3, -1, 10)}
cu_clf = cumlRidge(alpha=alpha,
fit_intercept=fit_intercept,
normalize=normalize,
solver="eig")
assert getattr(cu_clf, 'score', False)
sk_cu_grid = GridSearchCV(cu_clf, params, cv=5, iid=False)
gdf_data = cudf.DataFrame(X_train)
gdf_train = cudf.DataFrame(dict(train=y_train))
sk_cu_grid.fit(gdf_data, gdf_train.train)
assert sk_cu_grid.best_params_ == {'alpha': 0.1}
def make_regression_dataset(datatype, nrows, ncols):
ninformative = max(min(ncols, 5), int(math.ceil(ncols / 5)))
X, y = data.make_regression(dtype=datatype,
n_samples=nrows + 1000,
n_features=ncols,
random_state=SEED,
n_informative=ninformative)
return dsel.train_test_split(X, y, random_state=SEED, train_size=nrows)
def test_split_df_single_argument(test_size, train_size, shuffle):
X = cudf.DataFrame({'x': range(50)})
X_train, X_test = train_test_split(X,
train_size=train_size,
test_size=test_size,
shuffle=shuffle,
random_state=0)
if train_size is not None:
assert X_train.shape[0] == (int)(X.shape[0] * train_size)
if test_size is not None:
assert X_test.shape[0] == (int)(X.shape[0] * test_size)
def load_data(fpath):
"""
Simple helper function for loading data to be used by CPU/GPU models.
:param fpath: Path to the data to be ingested
:return: DataFrame wrapping the data at [fpath]. Data will be in either a Pandas or RAPIDS (cuDF) DataFrame
"""
import cudf
df = cudf.read_parquet(fpath)
X = df.drop(["ArrDelayBinary"], axis=1)
y = df["ArrDelayBinary"].astype("int32")
return train_test_split(X, y, test_size=0.2)
def test_stratified_random_seed(seed_type):
for i in range(10):
seed_n = np.random.randint(0, int(1e9))
if seed_type == 'int':
seed = seed_n
if seed_type == 'cupy':
seed = cp.random.RandomState(seed=seed_n)
if seed_type == 'numpy':
seed = np.random.RandomState(seed=seed_n)
X = cudf.DataFrame({"x": range(100)})
y = cudf.Series(([0] * (100 // 2)) + ([1] * (100 // 2)))
X_train, X_test, y_train, y_test = train_test_split(X,
y,
random_state=seed,
stratify=y)
if seed_type == 'cupy':
seed = cp.random.RandomState(seed=seed_n)
if seed_type == 'numpy':
seed = np.random.RandomState(seed=seed_n)
X_train2, X_test2, y_train2, y_test2 = \
train_test_split(X, y, random_state=seed, stratify=y)
assert X_train.equals(X_train2)
assert X_test.equals(X_test2)
assert y_train.equals(y_train2)
assert y_test.equals(y_test2)
# Ensure that data is shuffled
assert not (X.head().index.values == X_train.head().index.values).all()
def monotonic_inc(x):
dx = cp.diff(x.values, axis=0)
return cp.all(dx == 1)
assert not monotonic_inc(X_train)
def _preprocess_data(self, train_data, labels, batch_size, train_size,
truncate):
train_gdf = cudf.DataFrame()
train_gdf["domain"] = train_data
train_gdf["type"] = labels
domain_train, domain_test, type_train, type_test = train_test_split(
train_gdf, "type", train_size=train_size)
test_df = self._create_df(domain_test, type_test)
train_df = self._create_df(domain_train, type_train)
test_dataset = DGADataset(test_df, truncate)
train_dataset = DGADataset(train_df, truncate)
test_dataloader = DataLoader(test_dataset, batchsize=batch_size)
train_dataloader = DataLoader(train_dataset, batchsize=batch_size)
return train_dataloader, test_dataloader
def test_stratify_retain_index(test_size, train_size):
X = cudf.DataFrame({"x": range(10)})
y = cudf.Series(([0] * (10 // 2)) + ([1] * (10 // 2)))
X_train, X_test, y_train, y_test = train_test_split(X, y,
train_size=train_size,
test_size=test_size,
shuffle=True,
stratify=True)
assert (X_train["x"] == X_train.index).all()
assert (X_test["x"] == X_test.index).all()
if train_size is not None:
assert X_train.shape[0] == (int)(X.shape[0] * train_size)
elif test_size is not None:
assert X_test.shape[0] == (int)(X.shape[0] * test_size)
def split_dataset(self, dataset, random_state):
"""
Split dataset into train and test data subsets,
currently using CV-fold index for randomness.
Plan to refactor with sklearn KFold
"""
hpo_log.info('> train-test split')
label_column = self.hpo_config.label_column
X_train, X_test, y_train, y_test = \
train_test_split(dataset, label_column,
random_state=random_state)
return (X_train.astype(self.hpo_config.dataset_dtype),
X_test.astype(self.hpo_config.dataset_dtype),
y_train.astype(self.hpo_config.dataset_dtype),
y_test.astype(self.hpo_config.dataset_dtype))
def test_default_values():
X = np.zeros((100, 10)) + np.arange(100).reshape(100, 1)
y = np.arange(100).reshape(100, 1)
X = cp.asarray(X)
y = cp.asarray(y)
X_train, X_test, y_train, y_test = train_test_split(X, y)
assert isinstance(X_train, cp.ndarray)
assert isinstance(X_test, cp.ndarray)
assert isinstance(y_train, cp.ndarray)
assert isinstance(y_test, cp.ndarray)
assert X_train.shape[0] == X.shape[0] * 0.75
assert y_train.shape[0] == y.shape[0] * 0.75
assert X_test.shape[0] == X.shape[0] * 0.25
assert y_test.shape[0] == y.shape[0] * 0.25
def load_data(fpath):
"""
Simple helper function for loading data to be used by CPU/GPU models.
:param fpath: Path to the data to be ingested
:return: DataFrame wrapping the data at [fpath]. Data will be in either a Pandas or RAPIDS (cuDF) DataFrame
"""
import cudf
if (fpath.startswith('gs://')):
fs = gcsfs.GCSFileSystem()
with fs.open(fpath, mode='rb') as f:
df = cudf.read_parquet(f)
else:
df = cudf.read_parquet(fpath)
X = df.drop(["ArrDelayBinary"], axis=1)
y = df["ArrDelayBinary"].astype("int32")
return train_test_split(X, y, test_size=0.2)
def test_stratify_any_input(test_size, train_size):
X = cudf.DataFrame({"x": range(10)})
X['test_col'] = cudf.Series([10, 0, 0, 10, 10, 10, 0, 0, 10, 10])
y = cudf.Series(([0] * (10 // 2)) + ([1] * (10 // 2)))
X_train, X_test, y_train, y_test = train_test_split(X,
y,
train_size=train_size,
test_size=test_size,
shuffle=True,
stratify=X['test_col'],
random_state=15)
assert (X_train["x"].to_numpy() == X_train.index.to_numpy()).all()
assert (X_test["x"].to_numpy() == X_test.index.to_numpy()).all()
if train_size is not None:
assert X_train.shape[0] == (int)(X.shape[0] * train_size)
elif test_size is not None:
assert X_test.shape[0] == (int)(X.shape[0] * test_size)
def test_split_dataframe(train_size, shuffle):
X = cudf.DataFrame({"x": range(100)})
y = cudf.Series(([0] * (100 // 2)) + ([1] * (100 // 2)))
X_train, X_test, y_train, y_test = train_test_split(X,
y,
train_size=train_size,
shuffle=shuffle)
assert len(X_train) == len(y_train) == pytest.approx(train_size * len(X))
assert (len(X_test) == len(y_test) == pytest.approx(
(1 - train_size) * len(X)))
assert (all(X_train.index.to_pandas() == y_train.index.to_pandas()))
assert (all(X_test.index.to_pandas() == y_test.index.to_pandas()))
X_reconstructed = cudf.concat([X_train, X_test]).sort_values(by=["x"])
y_reconstructed = y_train.append(y_test).sort_values()
assert all(X_reconstructed.reset_index(drop=True) == X)
out = y_reconstructed.reset_index(drop=True).values_host == y.values_host
assert all(out)
def test_train_model():
if torch.cuda.is_available():
fake = Faker()
email_col = [fake.text() for _ in range(200)]
label_col = [random.randint(0, 1) for _ in range(200)]
emails_gdf = cudf.DataFrame(list(zip(email_col, label_col)), columns=["email", "label"])
X_train, X_test, y_train, y_test = train_test_split(
emails_gdf, "label", train_size=0.8, random_state=10
)
sc.train_model(
X_train["email"],
y_train,
learning_rate=3e-5,
max_seq_len=128,
batch_size=6,
epochs=1,
)
assert isinstance(
sc._model.module,
transformers.models.bert.modeling_bert.BertForSequenceClassification,
)
def test_stratified_split(type, test_size, train_size):
# For more tolerance and reliable estimates
X, y = make_classification(n_samples=10000)
if type == 'cupy':
X = cp.asarray(X)
y = cp.asarray(y)
if type == 'numba':
X = cuda.to_device(X)
y = cuda.to_device(y)
def counts(y):
_, y_indices = cp.unique(y, return_inverse=True)
class_counts = cp.bincount(y_indices)
total = cp.sum(class_counts)
percent_counts = []
for count in (class_counts):
percent_counts.append(
cp.around(float(count) / total.item(), decimals=2).item())
return percent_counts
X_train, X_test, y_train, y_test = train_test_split(X,
y,
train_size=train_size,
test_size=test_size,
stratify=y)
original_counts = counts(y)
split_counts = counts(y_train)
assert cp.isclose(original_counts, split_counts, equal_nan=False,
rtol=0.1).all()
if type == 'cupy':
assert isinstance(X_train, cp.ndarray)
assert isinstance(X_test, cp.ndarray)
if type in ['numba']:
assert cuda.devicearray.is_cuda_ndarray(X_train)
assert cuda.devicearray.is_cuda_ndarray(X_test)
def test_hinge_loss(nrows, ncols, n_info, input_type, n_classes):
train_rows = np.int32(nrows * 0.8)
X, y = make_classification(n_samples=nrows,
n_features=ncols,
n_clusters_per_class=1,
n_informative=n_info,
random_state=123,
n_classes=n_classes)
if input_type == "cudf":
X = cudf.DataFrame(X)
y = cudf.Series(y)
elif input_type == "cupy":
X = cp.asarray(X)
y = cp.asarray(y)
X_train, X_test, y_train, y_test = train_test_split(X,
y,
train_size=train_rows,
shuffle=True)
cuml_model = cu_log()
cuml_model.fit(X_train, y_train)
cu_predict_decision = cuml_model.decision_function(X_test)
cu_loss = cuml_hinge(y_test, cu_predict_decision.T, labels=cp.unique(y))
if input_type == "cudf":
y_test = y_test.to_array()
y = y.to_array()
cu_predict_decision = cp.asnumpy(cu_predict_decision.values)
elif input_type == "cupy":
y = cp.asnumpy(y)
y_test = cp.asnumpy(y_test)
cu_predict_decision = cp.asnumpy(cu_predict_decision)
cu_loss_using_sk = sk_hinge(y_test,
cu_predict_decision.T,
labels=np.unique(y))
# compare the accuracy of the two models
cp.testing.assert_array_almost_equal(cu_loss, cu_loss_using_sk)
import joblib
import os
import cuml
from cuml.benchmark.datagen import load_higgs
OUTPUT_DIR = './outputs/'
os.makedirs(OUTPUT_DIR, exist_ok=True)
X, y = load_higgs()
N_ROWS = 1000000
run = Run.get_context()
client = ExplanationClient.from_run(run)
run.log('N_ROWS', N_ROWS)
X_train, X_test, y_train, y_test = train_test_split(X[:N_ROWS],
y[:N_ROWS],
random_state=1)
# write x_test out as a pickle file for later visualization
x_test_pkl = 'x_test.pkl'
with open(x_test_pkl, 'wb') as file:
joblib.dump(value=X_test, filename=os.path.join(OUTPUT_DIR, x_test_pkl))
run.upload_file('x_test_higgs.pkl', os.path.join(OUTPUT_DIR, x_test_pkl))
gamma = 0.001
C = 100.
# Use SVC algorithm to create a model
reg = cuml.svm.SVC(C=C, gamma=gamma, probability=True)
model = reg.fit(X_train, y_train)
# preds = reg.predict(X_test)
def train_model(self,
train_gdf,
cat_cols,
cont_cols,
label_col,
batch_size,
epochs,
lr=0.01,
wd=0.0):
"""
This function is used for training fastai tabular model with a given training dataset.
:param train_gdf: training dataset with categorized and/or continuous feature columns
:type train_gdf: cudf.DataFrame
:param cat_cols: array of categorical column names in train_gdf
:type label_col: array
:param cont_col: array of continuous column names in train_gdf
:type label_col: array
:param label_col: column name of label column in train_gdf
:type label_col: str
:param batch_size: train_gdf will be partitioned into multiple dataframes of this size
:type batch_size: int
:param epochs: number of epochs to be adjusted depending on convergence for a specific dataset
:type epochs: int
:param lr: learning rate
:type lr: float
:param wd: wd
:type wd: float
Examples
--------
>>> from clx.analytics.asset_classification import AssetClassification
>>> ac = AssetClassification()
>>> cat_cols = ["1", "2", "3", "4", "5", "6", "7", "8", "9"]
>>> cont_cols = ["10"]
>>> ac.train_model(X_train, cat_cols, cont_cols, "label", batch_size, epochs, lr=0.01, wd=0.0)
"""
self._cat_cols = cat_cols
self._cont_cols = cont_cols
# train/test split
X, val_X, Y, val_Y = train_test_split(train_gdf,
label_col,
train_size=0.9)
val_X.index = val_Y.index
X.index = Y.index
embedded_cols = {}
for col in cat_cols:
if col != label_col:
categories_cnt = X[col].max() + 2
if categories_cnt > 1:
embedded_cols[col] = categories_cnt
X[label_col] = Y
val_X[label_col] = val_Y
# Embedding
embedding_sizes = [(n_categories, min(100, (n_categories + 1) // 2))
for _, n_categories in embedded_cols.items()]
n_cont = len(cont_cols)
out_sz = train_gdf[label_col].nunique()
# Partition dataframes
train_part_dfs = self._get_partitioned_dfs(X, batch_size)
val_part_dfs = self._get_partitioned_dfs(val_X, batch_size)
self._model = TabularModel(embedding_sizes, n_cont, out_sz,
self._layers, self._drops, self._emb_drop,
self._is_reg, self._is_multi, self._use_bn)
self._to_device(self._model, self._device)
self._config_optimizer()
for i in range(epochs):
loss = self._train(self._model, self._optimizer, train_part_dfs,
cat_cols, cont_cols, label_col)
print("training loss: ", loss)
self._val_loss(self._model, val_part_dfs, cat_cols, cont_cols,
label_col)
def classification_dataset(request):
X, y = make_classification(n_samples=10, n_features=5, random_state=0)
return train_test_split(X, y, random_state=0)
def test_split_invalid_proportion(train_size):
X = cudf.DataFrame({'x': range(10)})
y = cudf.Series([0] * 10)
with pytest.raises(ValueError):
train_test_split(X, y, train_size=train_size)
def test_split_size_mismatch():
X = cudf.DataFrame({'x': range(3)})
y = cudf.Series([0, 1])
with pytest.raises(ValueError):
train_test_split(X, y)
def test_qn(loss, dtype, penalty, l1_strength, l2_strength, fit_intercept):
if penalty == "none" and (l1_strength > 0 or l2_strength > 0):
pytest.skip("`none` penalty does not take l1/l2_strength")
tol = 1e-6
qn = cuQN(loss=loss,
fit_intercept=fit_intercept,
l1_strength=l1_strength,
l2_strength=l2_strength,
tol=1e-8,
output_type="cupy")
if loss == 'softmax':
X, y = make_classification(n_samples=5000,
n_informative=10,
n_features=20,
n_classes=4,
dtype=dtype)
stratify = y.astype(dtype)
X_train, X_test, y_train, y_test = train_test_split(X.astype(dtype),
y.astype(dtype),
stratify=stratify)
most_class = cp.unique(y)[cp.argmax(cp.bincount(y))]
baseline_preds = cp.array([most_class] * y_test.shape[0], dtype=dtype)
baseline_score = accuracy_score(y_test, baseline_preds)
y_pred = qn.fit(X_train, y_train).predict(X_test)
cuml_score = accuracy_score(y_test, y_pred)
assert (cuml_score > baseline_score)
assert (cuml_score >= 0.50)
elif loss == 'sigmoid':
X = np.array(precomputed_X, dtype=dtype)
y = np.array(precomputed_y_log, dtype=dtype)
qn.fit(X, y)
print(qn.objective)
print(qn.coef_)
if penalty == 'none' and l1_strength == 0.0 and l2_strength == 0.0:
if fit_intercept:
assert (qn.objective - 0.40263831615448) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array([[-2.1088872],
[2.4812558]]),
decimal=3)
else:
assert (qn.objective - 0.4317452311515808) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array([[-2.120777],
[3.056865]]),
decimal=3)
elif penalty == 'l1' and l2_strength == 0.0:
if fit_intercept:
if l1_strength == 0.0:
assert (qn.objective - 0.40263831615448) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array(
[[-2.1088872],
[2.4812558]]),
decimal=3)
else:
assert (qn.objective - 0.44295936822891235) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array(
[[-1.6899368],
[1.9021575]]),
decimal=3)
else:
if l1_strength == 0.0:
assert (qn.objective - 0.4317452311515808) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array([[-2.120777],
[3.056865]
]),
decimal=3)
else:
assert (qn.objective - 0.4769895672798157) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array(
[[-1.6214856],
[2.3650239]]),
decimal=3)
# assert False
elif penalty == 'l2' and l1_strength == 0.0:
if fit_intercept:
if l2_strength == 0.0:
assert (qn.objective - 0.40263831615448) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array(
[[-2.1088872],
[2.4812558]]),
decimal=3)
else:
assert (qn.objective - 0.43780848383903503) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array(
[[-1.5337948],
[1.678699]]),
decimal=3)
else:
if l2_strength == 0.0:
assert (qn.objective - 0.4317452311515808) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array([[-2.120777],
[3.056865]
]),
decimal=3)
else:
assert (qn.objective - 0.4750209450721741) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array(
[[-1.3931049],
[2.0140104]]),
decimal=3)
if penalty == 'elasticnet':
if fit_intercept:
if l1_strength == 0.0 and l2_strength == 0.0:
assert (qn.objective - 0.40263831615448) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array(
[[-2.1088872],
[2.4812558]]),
decimal=3)
elif l1_strength == 0.0:
assert (qn.objective - 0.43780848383903503) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array(
[[-1.5337948],
[1.678699]]),
decimal=3)
elif l2_strength == 0.0:
assert (qn.objective - 0.44295936822891235) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array(
[[-1.6899368],
[1.9021575]]),
decimal=3)
else:
assert (qn.objective - 0.467987984418869) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array(
[[-1.3727235],
[1.4639963]]),
decimal=3)
else:
if l1_strength == 0.0 and l2_strength == 0.0:
assert (qn.objective - 0.4317452311515808) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array([[-2.120777],
[3.056865]
]),
decimal=3)
elif l1_strength == 0.0:
assert (qn.objective - 0.4750209450721741) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array(
[[-1.3931049],
[2.0140104]]),
decimal=3)
elif l2_strength == 0.0:
assert (qn.objective - 0.4769895672798157) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array(
[[-1.6214856],
[2.3650239]]),
decimal=3)
else:
assert (qn.objective - 0.5067970156669617) < tol
cp.testing.assert_array_almost_equal(qn.coef_,
np.array(
[[-1.2102532],
[1.752459]]),
decimal=3)
print()