def test_sum_kernel_grad():
npr.seed(1)
eps = 1e-5
N = 10
M = 5
D = 3
kernel1 = Matern52(D)
kernel2 = Matern52(D)
kernel3 = Matern52(D)
kernel = SumKernel(kernel1, kernel2, kernel3)
data1 = npr.randn(N, D)
data2 = npr.randn(M, D)
loss = np.sum(kernel.cross_cov(data1, data2))
dloss = kernel.cross_cov_grad_data(data1, data2).sum(0)
dloss_est = np.zeros(dloss.shape)
for i in xrange(M):
for j in xrange(D):
data2[i, j] += eps
loss_1 = np.sum(kernel.cross_cov(data1, data2))
data2[i, j] -= 2 * eps
loss_2 = np.sum(kernel.cross_cov(data1, data2))
data2[i, j] += eps
dloss_est[i, j] = ((loss_1 - loss_2) / (2 * eps))
assert np.linalg.norm(dloss - dloss_est) < 1e-6
def test_grad():
npr.seed(1)
eps = 1e-5
N = 10
M = 5
D = 5
inds = [0, 2, 4]
kernel = Subset(D, Matern52(len(inds)), inds)
data1 = npr.randn(N, D)
data2 = npr.randn(M, D)
loss = np.sum(kernel.cross_cov(data1, data2))
dloss = kernel.cross_cov_grad_data(data1, data2).sum(0)
dloss_est = np.zeros(dloss.shape)
for i in range(M):
for j in range(D):
data2[i, j] += eps
loss_1 = np.sum(kernel.cross_cov(data1, data2))
data2[i, j] -= 2 * eps
loss_2 = np.sum(kernel.cross_cov(data1, data2))
data2[i, j] += eps
dloss_est[i, j] = (old_div((loss_1 - loss_2), (2 * eps)))
print('Subset kernel grad using indices %s:' % inds)
print(dloss)
assert np.linalg.norm(dloss - dloss_est) < 1e-6
def test_grad():
npr.seed(1)
eps = 1e-5
N = 10
M = 5
D = 3
kernel = Scale(Matern52(D))
kernel.amp2.value = 5.75
data1 = npr.randn(N, D)
data2 = npr.randn(M, D)
loss = np.sum(kernel.cross_cov(data1, data2))
dloss = kernel.cross_cov_grad_data(data1, data2).sum(0)
dloss_est = np.zeros(dloss.shape)
for i in range(M):
for j in range(D):
data2[i, j] += eps
loss_1 = np.sum(kernel.cross_cov(data1, data2))
data2[i, j] -= 2 * eps
loss_2 = np.sum(kernel.cross_cov(data1, data2))
data2[i, j] += eps
dloss_est[i, j] = ((loss_1 - loss_2) / (2 * eps))
assert np.linalg.norm(dloss - dloss_est) < 1e-6
def _build(self):
self.params = {}
self.latent_values = None
# Build the transformer
beta_warp = BetaWarp(self.num_dims)
beta_alpha, beta_beta = beta_warp.hypers
self.params['beta_alpha'] = beta_alpha
self.params['beta_beta'] = beta_beta
transformer = Transformer(self.num_dims)
transformer.add_layer(beta_warp)
# Build the component kernels
input_kernel = Matern52(self.num_dims)
ls = input_kernel.hypers
self.params['ls'] = ls
# Now apply the transformation.
transform_kernel = TransformKernel(input_kernel, transformer)
# Add some perturbation for stability
stability_noise = Noise(self.num_dims)
# Finally make a noisy version if necessary
# In a classifier GP the notion of "noise" is really just the scale.
if self.noiseless:
self._kernel = SumKernel(transform_kernel, stability_noise)
else:
scaled_kernel = Scale(transform_kernel)
self._kernel = SumKernel(scaled_kernel, stability_noise)
amp2 = scaled_kernel.hypers
self.params['amp2'] = amp2
# Build the mean function (just a constant mean for now)
self.mean = Hyperparameter(initial_value=0.0,
prior=priors.Gaussian(0.0, 1.0),
name='mean')
self.params['mean'] = self.mean
# Buld the latent values. Empty for now until the GP gets data.
self.latent_values = Hyperparameter(initial_value=np.array([]),
name='latent values')
# Build the samplers
to_sample = [self.mean] if self.noiseless else [self.mean, amp2]
self._samplers.append(
SliceSampler(*to_sample, compwise=False, thinning=self.thinning))
self._samplers.append(
WhitenedPriorSliceSampler(ls,
beta_alpha,
beta_beta,
compwise=True,
thinning=self.thinning))
self.latent_values_sampler = EllipticalSliceSampler(
self.latent_values, thinning=self.ess_thinning)
def test_grad():
npr.seed(1)
eps = 1e-5
N = 10
M = 5
D = 5
beta_warp = BetaWarp(2)
norm = Normalization(2)
lin = Linear(D)
transformer = Transformer(D)
# Each entry is a tuple, (transformation, indices_it_acts_on)
transformer.add_layer(
(beta_warp, [0, 2]),
(norm, [1, 4])) # This is crazy. We would never do this.
# One transformation means apply to all dimensions.
transformer.add_layer(lin)
kernel = TransformKernel(Matern52(lin.num_factors), transformer)
data1 = npr.rand(N, D)
data2 = npr.rand(M, D)
loss = np.sum(kernel.cross_cov(data1, data2))
dloss = kernel.cross_cov_grad_data(data1, data2).sum(0)
dloss_est = np.zeros(dloss.shape)
for i in xrange(M):
for j in xrange(D):
data2[i, j] += eps
loss_1 = np.sum(kernel.cross_cov(data1, data2))
data2[i, j] -= 2 * eps
loss_2 = np.sum(kernel.cross_cov(data1, data2))
data2[i, j] += eps
dloss_est[i, j] = ((loss_1 - loss_2) / (2 * eps))
assert np.linalg.norm(dloss - dloss_est) < 1e-6