def test_update_grid(engine, N=100, D=1, seed=42):
gp = GP_BACKEND(engine)
rng = np.random.RandomState(seed)
X = rng.normal(size=(N, D))
# Test updating the grid
gp.configure(**msgp_test_config)
gp.update_data('tmp', X)
gp.update_grid('tmp')
gp.eng.eval("assert(exist('xg','var') ~= 0)")
def test_update_data(engine, N=100, D=1, seed=42):
gp = GP_BACKEND(engine)
rng = np.random.RandomState(seed)
X = rng.normal(size=(N, D))
y = rng.normal(size=(N, D))
# Test updating data
gp.configure(**gp_test_config)
gp.update_data('tmp', X, y)
gp.eng.eval("assert(exist('X_tmp','var') ~= 0)")
gp.eng.eval("assert(exist('y_tmp','var') ~= 0)")
def test_evaluate(engine, N=100, D=1, seed=42):
gp = GP_BACKEND(engine)
rng = np.random.RandomState(seed)
X = rng.normal(size=(N, D))
y = rng.normal(size=(N, D))
gp.update_data('tmp', X, y)
# Evaluate GP
gp.configure(**gp_test_config)
nlZ = gp.evaluate('tmp')
# Evaluate MSGP
gp.configure(**msgp_test_config)
gp.update_grid('tmp')
nlZ = gp.evaluate('tmp')
def test_get_dlik_dx(engine, N=100, D=1, seed=42):
gp = GP_BACKEND(engine)
rng = np.random.RandomState(seed)
X = rng.normal(size=(N, D))
y = rng.normal(size=(N, D))
gp.update_data('tmp', X, y)
# GP
gp.configure(**gp_test_config)
dlik_dx = gp.get_dlik_dx('tmp')
assert type(dlik_dx) is np.ndarray
assert dlik_dx.shape == (N, D)
# MSGP
gp.configure(**msgp_test_config)
gp.update_grid('tmp')
dlik_dx = gp.get_dlik_dx('tmp')
assert type(dlik_dx) is np.ndarray
assert dlik_dx.shape == (N, D)
def test_train(engine, N=100, D=1, seed=42):
gp = GP_BACKEND(engine)
rng = np.random.RandomState(seed)
X_tr = rng.normal(size=(N, D))
y_tr = rng.normal(size=(N, D))
# Train a GP
gp.configure(**gp_test_config)
hyp = gp.train(5, X_tr, y_tr)
assert type(hyp) is dict
assert set(hyp.keys()) == set(gp_test_config['hyp'].keys())
# Train an MSGP
gp.configure(**msgp_test_config)
gp.update_data('tr', X_tr, y_tr)
gp.update_grid('tr')
hyp = gp.train(5)
assert type(hyp) is dict
assert set(hyp.keys()) == set(gp_test_config['hyp'].keys())
def test_predict(engine, N=100, D=1, seed=42):
gp = GP_BACKEND(engine)
rng = np.random.RandomState(seed)
X_tr = rng.normal(size=(N, D))
y_tr = rng.normal(size=(N, D))
X_tst = rng.normal(size=(N, D))
gp.update_data('tr', X_tr, y_tr)
# Predict using GP
gp.configure(**gp_test_config)
ymu, ys2 = gp.predict(X_tst, return_var=True)
assert type(ymu) is np.ndarray and type(ys2) is np.ndarray
assert ymu.shape == ys2.shape
# Predict using MSGP
gp.configure(**msgp_test_config)
gp.update_grid('tr')
ymu, ys2 = gp.predict(X_tst, return_var=True)
assert type(ymu) is np.ndarray and type(ys2) is np.ndarray
assert ymu.shape == ys2.shape