apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-06-08 15:05:15 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

lots of bugfixes after refactoring

Browse source
This commit is contained in:
James Hensman 2013-06-05 16:14:43 +01:00
parent 3fbd7e4943
commit 527586a012
12 changed files with 53 additions and 53 deletions
Download patch file Download diff file Expand all files Collapse all files

8
GPy/core/fitc.py
View file

@ -20,7 +20,7 @@ class FITC(SparseGP):
sparse FITC approximation
:param X: inputs
:type X: np.ndarray (N x Q)
:type X: np.ndarray (num_data x Q)
:param likelihood: a likelihood instance, containing the observed data
:type likelihood: GPy.likelihood.(Gaussian | EP)
:param kernel : the kernel (covariance function). See link kernels
@ -112,7 +112,7 @@ class FITC(SparseGP):
else:
if self.likelihood.is_heteroscedastic:
assert self.likelihood.output_dim == 1
tmp = self.psi1 * (np.sqrt(self.beta_star.flatten().reshape(1, self.N)))
tmp = self.psi1 * (np.sqrt(self.beta_star.flatten().reshape(1, self.num_data)))
tmp, _ = linalg.lapack.flapack.dtrtrs(self.Lm, np.asfortranarray(tmp), lower=1)
self.A = tdot(tmp)
@ -168,7 +168,7 @@ class FITC(SparseGP):
self._dpsi1_dX_jkj = 0
self._dpsi1_dtheta_jkj = 0
for i,V_n,alpha_n,gamma_n,gamma_k in zip(range(self.N),self.V_star,alpha,gamma_2,gamma_3):
for i,V_n,alpha_n,gamma_n,gamma_k in zip(range(self.num_data),self.V_star,alpha,gamma_2,gamma_3):
K_pp_K = np.dot(Kmmipsi1[:,i:(i+1)],Kmmipsi1[:,i:(i+1)].T)
#Diag_dpsi1 = Diag_dA_dpsi1: yT*beta_star*y + Diag_dC_dpsi1 +Diag_dD_dpsi1
@ -215,7 +215,7 @@ class FITC(SparseGP):
def log_likelihood(self):
""" Compute the (lower bound on the) log marginal likelihood """
A = -0.5 * self.N * self.output_dim * np.log(2.*np.pi) + 0.5 * np.sum(np.log(self.beta_star)) - 0.5 * np.sum(self.V_star * self.likelihood.Y)
A = -0.5 * self.num_data * self.output_dim * np.log(2.*np.pi) + 0.5 * np.sum(np.log(self.beta_star)) - 0.5 * np.sum(self.V_star * self.likelihood.Y)
C = -self.output_dim * (np.sum(np.log(np.diag(self.LB))))
D = 0.5 * np.sum(np.square(self._LBi_Lmi_psi1V))
return A + C + D

4
GPy/core/gp.py
View file

@ -46,12 +46,12 @@ class GP(GPBase):
#alpha = np.dot(self.Ki, self.likelihood.Y)
alpha,_ = linalg.lapack.flapack.dpotrs(self.L, self.likelihood.Y,lower=1)
self.dL_dK = 0.5 * (tdot(alpha) - self.input_dim * self.Ki)
self.dL_dK = 0.5 * (tdot(alpha) - self.output_dim * self.Ki)
else:
#tmp = mdot(self.Ki, self.likelihood.YYT, self.Ki)
tmp, _ = linalg.lapack.flapack.dpotrs(self.L, np.asfortranarray(self.likelihood.YYT), lower=1)
tmp, _ = linalg.lapack.flapack.dpotrs(self.L, np.asfortranarray(tmp.T), lower=1)
self.dL_dK = 0.5 * (tmp - self.input_dim * self.Ki)
self.dL_dK = 0.5 * (tmp - self.output_dim * self.Ki)
def _get_params(self):
return np.hstack((self.kern._get_params_transformed(), self.likelihood._get_params()))

4
GPy/core/gp_base.py
View file

@ -13,12 +13,12 @@ class GPBase(model.model):
def __init__(self, X, likelihood, kernel, normalize_X=False):
self.X = X
assert len(self.X.shape) == 2
self.N, self.input_dim = self.X.shape
self.num_data, self.input_dim = self.X.shape
assert isinstance(kernel, kern.kern)
self.kern = kernel
self.likelihood = likelihood
assert self.X.shape[0] == self.likelihood.data.shape[0]
self.N, self.output_dim = self.likelihood.data.shape
self.num_data, self.output_dim = self.likelihood.data.shape
if normalize_X:
self._Xmean = X.mean(0)[None, :]

32
GPy/core/sparse_gp.py
View file

@ -13,13 +13,13 @@ class SparseGP(GPBase):
Variational sparse GP model
:param X: inputs
:type X: np.ndarray (N x input_dim)
:type X: np.ndarray (num_data x input_dim)
:param likelihood: a likelihood instance, containing the observed data
:type likelihood: GPy.likelihood.(Gaussian | EP | Laplace)
:param kernel : the kernel (covariance function). See link kernels
:type kernel: a GPy.kern.kern instance
:param X_variance: The uncertainty in the measurements of X (Gaussian variance)
:type X_variance: np.ndarray (N x input_dim) | None
:type X_variance: np.ndarray (num_data x input_dim) | None
:param Z: inducing inputs (optional, see note)
:type Z: np.ndarray (num_inducing x input_dim) | None
:param num_inducing : Number of inducing points (optional, default 10. Ignored if Z is not None)
@ -69,7 +69,7 @@ class SparseGP(GPBase):
# The rather complex computations of self.A
if self.has_uncertain_inputs:
if self.likelihood.is_heteroscedastic:
psi2_beta = (self.psi2 * (self.likelihood.precision.flatten().reshape(self.N, 1, 1))).sum(0)
psi2_beta = (self.psi2 * (self.likelihood.precision.flatten().reshape(self.num_data, 1, 1))).sum(0)
else:
psi2_beta = self.psi2.sum(0) * self.likelihood.precision
evals, evecs = linalg.eigh(psi2_beta)
@ -77,7 +77,7 @@ class SparseGP(GPBase):
tmp = evecs * np.sqrt(clipped_evals)
else:
if self.likelihood.is_heteroscedastic:
tmp = self.psi1 * (np.sqrt(self.likelihood.precision.flatten().reshape(1, self.N)))
tmp = self.psi1 * (np.sqrt(self.likelihood.precision.flatten().reshape(1, self.num_data)))
else:
tmp = self.psi1 * (np.sqrt(self.likelihood.precision))
tmp, _ = linalg.lapack.flapack.dtrtrs(self.Lm, np.asfortranarray(tmp), lower=1)
@ -99,28 +99,28 @@ class SparseGP(GPBase):
# Compute dL_dKmm
tmp = tdot(self._LBi_Lmi_psi1V)
self.DBi_plus_BiPBi = backsub_both_sides(self.LB, self.input_dim * np.eye(self.num_inducing) + tmp)
self.DBi_plus_BiPBi = backsub_both_sides(self.LB, self.output_dim * np.eye(self.num_inducing) + tmp)
tmp = -0.5 * self.DBi_plus_BiPBi
tmp += -0.5 * self.B * self.input_dim
tmp += self.input_dim * np.eye(self.num_inducing)
tmp += -0.5 * self.B * self.output_dim
tmp += self.output_dim * np.eye(self.num_inducing)
self.dL_dKmm = backsub_both_sides(self.Lm, tmp)
# Compute dL_dpsi # FIXME: this is untested for the heterscedastic + uncertain inputs case
self.dL_dpsi0 = -0.5 * self.input_dim * (self.likelihood.precision * np.ones([self.N, 1])).flatten()
self.dL_dpsi0 = -0.5 * self.output_dim * (self.likelihood.precision * np.ones([self.num_data, 1])).flatten()
self.dL_dpsi1 = np.dot(self.Cpsi1V, self.likelihood.V.T)
dL_dpsi2_beta = 0.5 * backsub_both_sides(self.Lm, self.input_dim * np.eye(self.num_inducing) - self.DBi_plus_BiPBi)
dL_dpsi2_beta = 0.5 * backsub_both_sides(self.Lm, self.output_dim * np.eye(self.num_inducing) - self.DBi_plus_BiPBi)
if self.likelihood.is_heteroscedastic:
if self.has_uncertain_inputs:
self.dL_dpsi2 = self.likelihood.precision.flatten()[:, None, None] * dL_dpsi2_beta[None, :, :]
else:
self.dL_dpsi1 += 2.*np.dot(dL_dpsi2_beta, self.psi1 * self.likelihood.precision.reshape(1, self.N))
self.dL_dpsi1 += 2.*np.dot(dL_dpsi2_beta, self.psi1 * self.likelihood.precision.reshape(1, self.num_data))
self.dL_dpsi2 = None
else:
dL_dpsi2 = self.likelihood.precision * dL_dpsi2_beta
if self.has_uncertain_inputs:
# repeat for each of the N psi_2 matrices
self.dL_dpsi2 = np.repeat(dL_dpsi2[None, :, :], self.N, axis=0)
self.dL_dpsi2 = np.repeat(dL_dpsi2[None, :, :], self.num_data, axis=0)
else:
# subsume back into psi1 (==Kmn)
self.dL_dpsi1 += 2.*np.dot(dL_dpsi2, self.psi1)
@ -135,24 +135,24 @@ class SparseGP(GPBase):
raise NotImplementedError, "heteroscedatic derivates not implemented"
else:
# likelihood is not heterscedatic
self.partial_for_likelihood = -0.5 * self.N * self.input_dim * self.likelihood.precision + 0.5 * self.likelihood.trYYT * self.likelihood.precision ** 2
self.partial_for_likelihood += 0.5 * self.input_dim * (self.psi0.sum() * self.likelihood.precision ** 2 - np.trace(self.A) * self.likelihood.precision)
self.partial_for_likelihood = -0.5 * self.num_data * self.output_dim * self.likelihood.precision + 0.5 * self.likelihood.trYYT * self.likelihood.precision ** 2
self.partial_for_likelihood += 0.5 * self.output_dim * (self.psi0.sum() * self.likelihood.precision ** 2 - np.trace(self.A) * self.likelihood.precision)
self.partial_for_likelihood += self.likelihood.precision * (0.5 * np.sum(self.A * self.DBi_plus_BiPBi) - np.sum(np.square(self._LBi_Lmi_psi1V)))
def log_likelihood(self):
""" Compute the (lower bound on the) log marginal likelihood """
if self.likelihood.is_heteroscedastic:
A = -0.5 * self.N * self.output_dim * np.log(2.*np.pi) + 0.5 * np.sum(np.log(self.likelihood.precision)) - 0.5 * np.sum(self.likelihood.V * self.likelihood.Y)
A = -0.5 * self.num_data * self.output_dim * np.log(2.*np.pi) + 0.5 * np.sum(np.log(self.likelihood.precision)) - 0.5 * np.sum(self.likelihood.V * self.likelihood.Y)
B = -0.5 * self.output_dim * (np.sum(self.likelihood.precision.flatten() * self.psi0) - np.trace(self.A))
else:
A = -0.5 * self.N * self.output_dim * (np.log(2.*np.pi) - np.log(self.likelihood.precision)) - 0.5 * self.likelihood.precision * self.likelihood.trYYT
A = -0.5 * self.num_data * self.output_dim * (np.log(2.*np.pi) - np.log(self.likelihood.precision)) - 0.5 * self.likelihood.precision * self.likelihood.trYYT
B = -0.5 * self.output_dim * (np.sum(self.likelihood.precision * self.psi0) - np.trace(self.A))
C = -self.output_dim * (np.sum(np.log(np.diag(self.LB)))) # + 0.5 * self.num_inducing * np.log(sf2))
D = 0.5 * np.sum(np.square(self._LBi_Lmi_psi1V))
return A + B + C + D + self.likelihood.Z
def _set_params(self, p):
self.Z = p[:self.num_inducing * self.output_dim].reshape(self.num_inducing, self.input_dim)
self.Z = p[:self.num_inducing * self.input_dim].reshape(self.num_inducing, self.input_dim)
self.kern._set_params(p[self.Z.size:self.Z.size + self.kern.Nparam])
self.likelihood._set_params(p[self.Z.size + self.kern.Nparam:])
self._compute_kernel_matrices()