def test_linear_regression_model(datatype, algorithm, nrows, column_info):
if algorithm == "svd" and nrows > 46340:
pytest.skip("svd solver is not supported for the data that has more"
"than 46340 rows or columns if you are using CUDA version"
"10.x")
ncols, n_info = column_info
X_train, X_test, y_train, y_test = make_regression_dataset(
datatype, nrows, ncols, n_info
)
# Initialization of cuML's linear regression model
cuols = cuLinearRegression(fit_intercept=True,
normalize=False,
algorithm=algorithm)
# fit and predict cuml linear regression model
cuols.fit(X_train, y_train)
cuols_predict = cuols.predict(X_test)
if nrows < 500000:
# sklearn linear regression model initialization, fit and predict
skols = skLinearRegression(fit_intercept=True, normalize=False)
skols.fit(X_train, y_train)
skols_predict = skols.predict(X_test)
assert array_equal(skols_predict, cuols_predict, 1e-1, with_sign=True)
def test_ols(datatype, X_type, y_type, algorithm):
X = np.array([[2.0, 5.0], [6.0, 9.0], [2.0, 2.0], [2.0, 3.0]],
dtype=datatype)
y = np.dot(X, np.array([5.0, 10.0]).astype(datatype))
pred_data = np.array([[3.0, 5.0], [2.0, 5.0]]).astype(datatype)
skols = skLinearRegression(fit_intercept=True, normalize=False)
skols.fit(X, y)
cuols = cuLinearRegression(fit_intercept=True,
normalize=False,
algorithm=algorithm)
if X_type == 'dataframe':
gdf = cudf.DataFrame()
gdf['0'] = np.asarray([2, 6, 2, 2], dtype=datatype)
gdf['1'] = np.asarray([5, 9, 2, 3], dtype=datatype)
cuols.fit(gdf, y)
elif X_type == 'ndarray':
cuols.fit(X, y)
sk_predict = skols.predict(pred_data)
cu_predict = cuols.predict(pred_data).to_array()
print(sk_predict)
print(cu_predict)
# print(skols.coef_)
print(cuols.gdf_datatype)
print(y.dtype)
assert array_equal(sk_predict, cu_predict, 1e-3, with_sign=True)
def test_weighted_linear_regression(datatype, algorithm, fit_intercept,
normalize, distribution):
nrows, ncols, n_info = 1000, 20, 10
max_weight = 10
noise = 20
X_train, X_test, y_train, y_test = make_regression_dataset(
datatype, nrows, ncols, n_info, noise=noise
)
# set weight per sample to be from 1 to max_weight
if distribution == "uniform":
wt = np.random.randint(1, high=max_weight, size=len(X_train))
elif distribution == "exponential":
wt = np.random.exponential(size=len(X_train))
else:
wt = np.random.lognormal(size=len(X_train))
# Initialization of cuML's linear regression model
cuols = cuLinearRegression(fit_intercept=fit_intercept,
normalize=normalize,
algorithm=algorithm)
# fit and predict cuml linear regression model
cuols.fit(X_train, y_train, sample_weight=wt)
cuols_predict = cuols.predict(X_test)
# sklearn linear regression model initialization, fit and predict
skols = skLinearRegression(fit_intercept=fit_intercept,
normalize=normalize)
skols.fit(X_train, y_train, sample_weight=wt)
skols_predict = skols.predict(X_test)
assert array_equal(skols_predict, cuols_predict, 1e-1, with_sign=True)
def test_linear_regression_model(datatype, algorithm, nrows, column_info):
ncols, n_info = column_info
X_train, X_test, y_train, y_test = make_regression_dataset(datatype,
nrows,
ncols,
n_info)
# Initialization of cuML's linear regression model
cuols = cuLinearRegression(fit_intercept=True,
normalize=False,
algorithm=algorithm)
# fit and predict cuml linear regression model
cuols.fit(X_train, y_train)
cuols_predict = cuols.predict(X_test)
if nrows < 500000:
# sklearn linear regression model initialization, fit and predict
skols = skLinearRegression(fit_intercept=True,
normalize=False)
skols.fit(X_train, y_train)
skols_predict = skols.predict(X_test)
assert array_equal(skols_predict, cuols_predict,
1e-1, with_sign=True)
def test_linreg_predict_convert_dtype(train_dtype, test_dtype):
X, y = make_regression(n_samples=50, n_features=10,
n_informative=5, random_state=0)
X = X.astype(train_dtype)
y = y.astype(train_dtype)
X_train, X_test, y_train, y_test = train_test_split(X, y, train_size=0.8,
random_state=0)
clf = cuLinearRegression()
clf.fit(X_train, y_train)
clf.predict(X_test.astype(test_dtype))
def getTheModel(ensemble, ntrees, frate, seed, coll):
seed = seed + 10
# if ntrees < 100:
# ntrees = 100
if frate < 0.3:
frate = 0.3
if ensemble == 1:
clf = RandomForestRegressor(n_estimators=ntrees,
max_features=frate,
max_leaf_nodes=100,
min_samples_leaf=1,
random_state=seed,
n_jobs=-1)
if ensemble == 2:
clf = ExtraTreesRegressor(n_estimators=ntrees,
max_leaf_nodes=100,
min_samples_leaf=1,
random_state=seed,
n_jobs=-1)
if ensemble == 3:
clf = GradientBoostingRegressor(n_estimators=ntrees,
learning_rate=0.1,
max_depth=2,
random_state=seed,
loss='ls')
if ensemble == 4:
clf = cuLinearRegression(fit_intercept=True,
normalize=True,
algorithm='eig')
# clf = linear_model.LinearRegression(n_jobs=-1)
if ensemble == 5:
clf = DecisionTreeRegressor(random_state=seed)
if ensemble == 10:
if "200" in coll:
clf = DecisionTreeRegressor(random_state=seed,
min_samples_leaf=30,
min_samples_split=50,
max_depth=3)
else:
clf = DecisionTreeRegressor(random_state=seed,
min_samples_leaf=50,
min_samples_split=100,
max_depth=6)
return clf
def test_linear_regression_model_default(datatype):
X_train, X_test, y_train, y_test = small_regression_dataset(datatype)
# Initialization of cuML's linear regression model
cuols = cuLinearRegression()
# fit and predict cuml linear regression model
cuols.fit(X_train, y_train)
cuols_predict = cuols.predict(X_test)
# sklearn linear regression model initialization and fit
skols = skLinearRegression()
skols.fit(X_train, y_train)
skols_predict = skols.predict(X_test)
assert array_equal(skols_predict, cuols_predict, 1e-1, with_sign=True)
def test_linear_regression_single_column():
'''Test that linear regression can be run on single column with more than
46340 rows (a limitation on CUDA <11)'''
model = cuLinearRegression()
model.fit(cp.random.rand(46341), cp.random.rand(46341))
def test_linear_models(datatype, X_type, y_type, algorithm, nrows, ncols,
n_info):
train_rows = np.int32(nrows * 0.8)
X, y = make_regression(n_samples=(nrows),
n_features=ncols,
n_informative=n_info,
random_state=0)
X_test = np.asarray(X[train_rows:, 0:]).astype(datatype)
X_train = np.asarray(X[0:train_rows, :]).astype(datatype)
y_train = np.asarray(y[0:train_rows, ]).astype(datatype)
# Initialization of cuML's linear and ridge regression models
cuols = cuLinearRegression(fit_intercept=True,
normalize=False,
algorithm=algorithm)
curidge = cuRidge(fit_intercept=False, normalize=False, solver=algorithm)
if X_type == 'dataframe':
y_train = pd.DataFrame({'labels': y_train[0:, ]})
X_train = pd.DataFrame(
{'fea%d' % i: X_train[0:, i]
for i in range(X_train.shape[1])})
X_test = pd.DataFrame(
{'fea%d' % i: X_test[0:, i]
for i in range(X_test.shape[1])})
X_cudf = cudf.DataFrame.from_pandas(X_train)
X_cudf_test = cudf.DataFrame.from_pandas(X_test)
y_cudf = y_train.values
y_cudf = y_cudf[:, 0]
y_cudf = cudf.Series(y_cudf)
# fit and predict cuml linear regression model
cuols.fit(X_cudf, y_cudf)
cuols_predict = cuols.predict(X_cudf_test).to_array()
# fit and predict cuml ridge regression model
curidge.fit(X_cudf, y_cudf)
curidge_predict = curidge.predict(X_cudf_test).to_array()
elif X_type == 'ndarray':
# fit and predict cuml linear regression model
cuols.fit(X_train, y_train)
cuols_predict = cuols.predict(X_test).to_array()
# fit and predict cuml ridge regression model
curidge.fit(X_train, y_train)
curidge_predict = curidge.predict(X_test).to_array()
if nrows < 500000:
# sklearn linear and ridge regression model initialization and fit
skols = skLinearRegression(fit_intercept=True, normalize=False)
skols.fit(X_train, y_train)
skridge = skRidge(fit_intercept=False, normalize=False)
skridge.fit(X_train, y_train)
skols_predict = skols.predict(X_test)
skridge_predict = skridge.predict(X_test)
assert array_equal(skols_predict, cuols_predict, 1e-1, with_sign=True)
assert array_equal(skridge_predict,
curidge_predict,
1e-1,
with_sign=True)
