def test_MixedKernelSeparateMof():
data = DataMixedKernel
kern_list = [SquaredExponential() for _ in range(data.L)]
inducing_variable_list = [
InducingPoints(data.X[:data.M, ...]) for _ in range(data.L)
]
k1 = mk.LinearCoregionalization(kern_list, W=data.W)
f1 = mf.SeparateIndependentInducingVariables(inducing_variable_list)
model_1 = SVGP(k1,
Gaussian(),
inducing_variable=f1,
q_mu=data.mu_data,
q_sqrt=data.sqrt_data)
kern_list = [SquaredExponential() for _ in range(data.L)]
inducing_variable_list = [
InducingPoints(data.X[:data.M, ...]) for _ in range(data.L)
]
k2 = mk.LinearCoregionalization(kern_list, W=data.W)
f2 = mf.SeparateIndependentInducingVariables(inducing_variable_list)
model_2 = SVGP(k2,
Gaussian(),
inducing_variable=f2,
q_mu=data.mu_data,
q_sqrt=data.sqrt_data)
check_equality_predictions(Data.data, [model_1, model_2])
def test_separate_independent_conditional_with_q_sqrt_none():
"""
In response to bug #1523, this test checks that separate_independent_condtional
does not fail when q_sqrt=None.
"""
q_sqrt = None
data = DataMixedKernel
kern_list = [SquaredExponential() for _ in range(data.L)]
kernel = gpflow.kernels.SeparateIndependent(kern_list)
inducing_variable_list = [
InducingPoints(data.X[:data.M, ...]) for _ in range(data.L)
]
inducing_variable = mf.SeparateIndependentInducingVariables(
inducing_variable_list)
mu_1, var_1 = gpflow.conditionals.conditional(
data.X,
inducing_variable,
kernel,
data.mu_data,
full_cov=False,
full_output_cov=False,
q_sqrt=q_sqrt,
white=True,
)
def test_separate_independent_mof():
"""
Same test as above but we use different (i.e. separate) inducing inducing
for each of the output dimensions.
"""
np.random.seed(0)
# Model 1 (INefficient)
q_mu_1 = np.random.randn(Data.M * Data.P, 1)
q_sqrt_1 = np.tril(np.random.randn(Data.M * Data.P,
Data.M * Data.P))[None,
...] # 1 x MP x MP
kernel_1 = mk.SharedIndependent(
SquaredExponential(variance=0.5, lengthscales=1.2), Data.P)
inducing_variable_1 = InducingPoints(Data.X[:Data.M, ...])
model_1 = SVGP(kernel_1,
Gaussian(),
inducing_variable_1,
q_mu=q_mu_1,
q_sqrt=q_sqrt_1)
set_trainable(model_1, False)
set_trainable(model_1.q_sqrt, True)
set_trainable(model_1.q_mu, True)
gpflow.optimizers.Scipy().minimize(
model_1.training_loss_closure(Data.data),
variables=model_1.trainable_variables,
method="BFGS",
compile=True,
)
# Model 2 (efficient)
q_mu_2 = np.random.randn(Data.M, Data.P)
q_sqrt_2 = np.array([
np.tril(np.random.randn(Data.M, Data.M)) for _ in range(Data.P)
]) # P x M x M
kernel_2 = mk.SharedIndependent(
SquaredExponential(variance=0.5, lengthscales=1.2), Data.P)
inducing_variable_list_2 = [
InducingPoints(Data.X[:Data.M, ...]) for _ in range(Data.P)
]
inducing_variable_2 = mf.SeparateIndependentInducingVariables(
inducing_variable_list_2)
model_2 = SVGP(kernel_2,
Gaussian(),
inducing_variable_2,
q_mu=q_mu_2,
q_sqrt=q_sqrt_2)
set_trainable(model_2, False)
set_trainable(model_2.q_sqrt, True)
set_trainable(model_2.q_mu, True)
gpflow.optimizers.Scipy().minimize(
model_2.training_loss_closure(Data.data),
variables=model_2.trainable_variables,
method="BFGS",
compile=True,
)
# Model 3 (Inefficient): an idenitical inducing variable is used P times,
# and treated as a separate one.
q_mu_3 = np.random.randn(Data.M, Data.P)
q_sqrt_3 = np.array([
np.tril(np.random.randn(Data.M, Data.M)) for _ in range(Data.P)
]) # P x M x M
kern_list = [
SquaredExponential(variance=0.5, lengthscales=1.2)
for _ in range(Data.P)
]
kernel_3 = mk.SeparateIndependent(kern_list)
inducing_variable_list_3 = [
InducingPoints(Data.X[:Data.M, ...]) for _ in range(Data.P)
]
inducing_variable_3 = mf.SeparateIndependentInducingVariables(
inducing_variable_list_3)
model_3 = SVGP(kernel_3,
Gaussian(),
inducing_variable_3,
q_mu=q_mu_3,
q_sqrt=q_sqrt_3)
set_trainable(model_3, False)
set_trainable(model_3.q_sqrt, True)
set_trainable(model_3.q_mu, True)
gpflow.optimizers.Scipy().minimize(
model_3.training_loss_closure(Data.data),
variables=model_3.trainable_variables,
method="BFGS",
compile=True,
)
check_equality_predictions(Data.data, [model_1, model_2, model_3])
class Datum:
D = 1
L = 2
P = 3
M = 10
N = 100
W = rng.randn(P, L)
X = rng.randn(N)[:, None]
Xnew = rng.randn(N)[:, None]
multioutput_inducing_variable_list = [
mf.SharedIndependentInducingVariables(make_ip()),
mf.SeparateIndependentInducingVariables(make_ips(Datum.P)),
]
multioutput_fallback_inducing_variable_list = [
mf.FallbackSharedIndependentInducingVariables(make_ip()),
mf.FallbackSeparateIndependentInducingVariables(make_ips(Datum.P)),
]
multioutput_kernel_list = [
mk.SharedIndependent(make_kernel(), Datum.P),
mk.SeparateIndependent(make_kernels(Datum.L)),
mk.LinearCoregionalization(make_kernels(Datum.L), Datum.W),
]
@pytest.mark.parametrize("inducing_variable",