def test_log_ML_from_lambda():
g = 0.0
model = gpmodel.LassoGPRegressor(gpkernel.LinearKernel())
model2 = gpmodel.GPRegressor(gpkernel.LinearKernel())
X, mask = model._regularize(X_df, gamma=g, y=Y)
model2.fit(X, Y)
neg_ML = model._log_ML_from_gamma(g, X, Y)
assert np.isclose(neg_ML, model2.ML)
def test_predict():
model = gpmodel.LassoGPRegressor(gpkernel.LinearKernel(), gamma=-2)
model.fit(X_df, Y)
np.random.seed(1)
X_test = np.random.random(size=(1, n_dims))
X_test = pd.DataFrame(X_test, index=['A'])
X_masked = X_test.transpose()[model._mask].transpose()
Y_test = model.predict(X_test)
check_model = gpmodel.GPRegressor(gpkernel.LinearKernel())
check_model.fit(model.X, Y)
Y_check = check_model.predict(X_masked)
assert np.isclose(Y_test, Y_check).all()
def test_linear_kernel():
kern = gpkernel.LinearKernel()
params = np.random.random(size=(1, ))
vp = params[0]
actual = X @ X.T * vp
assert np.allclose(kern.cov(X, X, params), actual)
assert np.allclose(kern.cov(xa, xb, params), actual[0, 1])
assert np.allclose(kern.cov(X[0:2], X[2:], params),
actual[0:2, 2:])
assert np.allclose(kern.cov(X, X), X @ X.T)
n = kern.fit(X)
assert n == 1
assert np.allclose(kern.cov(hypers=params), actual)
def test_regularize():
model = gpmodel.LassoGPRegressor(gpkernel.LinearKernel())
pytest.raises(ValueError, model._regularize, X_df, gamma=0)
X, mask = model._regularize(X_df, y=Y, gamma=0)
assert np.isclose(X.values,
X_df.transpose()[mask].transpose().values).all()
clf = linear_model.Lasso(alpha=np.exp(0))
clf.fit(X_df, Y)
not_zero = ~np.isclose(clf.coef_, 0)
assert np.array_equal(not_zero, mask)
new_mask = np.random.choice([True, False], size=len(X_df.columns))
X, mask = model._regularize(X_df, mask=new_mask)
assert np.array_equal(mask, new_mask)
assert np.isclose(X.values,
X_df.transpose()[mask].transpose().values).all()
from gpmodel import gpmodel
parser = argparse.ArgumentParser()
parser.add_argument('-k', '--kernel', required=True)
parser.add_argument('-type', '--type')
parser.add_argument('-g', '--hypers', nargs='*', required=False, type=float)
args = parser.parse_args()
if args.kernel == '3/2':
kern = gpkernel.MaternKernel(nu='3/2')
elif args.kernel == '5/2':
kern = gpkernel.MaternKernel(nu='5/2')
elif args.kernel == 'se':
kern = gpkernel.SEKernel()
elif args.kernel == 'linear':
kern = gpkernel.LinearKernel()
else:
raise ValueError('Invalid kernel.')
if args.type == 'b2':
np.random.seed(0)
cov = np.array([[1.2, 0.1], [0.1, 1.2]])
x1 = np.random.multivariate_normal(np.array([-3.0, -3.0]), cov, size=20)
x2 = np.random.multivariate_normal(np.array([1.0, 1.0]), cov, size=20)
x3 = np.random.multivariate_normal(np.array([0.0, 3.0]), cov, size=20)
x4 = np.random.multivariate_normal(np.array([4.0, -4.0]), cov, size=20)
X = np.concatenate((x1, x2, x3, x4))
X = pd.DataFrame(X)
X.index = [chr(i) for i in range(len(X))]
X.columns = ['x1', 'x2']
y1 = [1 for _ in range(40)]
def test_dump_and_load():
model = gpmodel.LassoGPRegressor(gpkernel.LinearKernel(), gamma=-2)
model.fit(X_df, Y)
model.dump('test.pkl')
model_2 = gpmodel.GPRegressor.load('test.pkl')
assert model.gamma == model_2.gamma
def test_fit():
model = gpmodel.LassoGPRegressor(gpkernel.LinearKernel(), gamma=0.1)
np.random.seed(1)
model.fit(X_df, Y)
assert model.X.shape[1] == 19
assert np.isclose(model.ML, 26.69784799760495)
def test_init():
model = gpmodel.LassoGPRegressor(gpkernel.LinearKernel(), gamma=1)
assert model._gamma_0 == 1
model = gpmodel.LassoGPRegressor(gpkernel.LinearKernel())
assert model._gamma_0 == 0