From 69f6cfa6f7671b32a77ff6bfdab4a114675abab5 Mon Sep 17 00:00:00 2001 From: Max Zwiessele Date: Fri, 11 Sep 2015 16:59:55 +0100 Subject: [PATCH] [inference] changed gaussian variance to precision (which it really is) --- .../latent_function_inference/dtc.py | 38 +++++++++---------- .../exact_gaussian_inference.py | 8 ++-- .../expectation_propagation.py | 4 +- .../expectation_propagation_dtc.py | 2 +- .../latent_function_inference/var_dtc.py | 38 +++++++++---------- 5 files changed, 45 insertions(+), 45 deletions(-) diff --git a/GPy/inference/latent_function_inference/dtc.py b/GPy/inference/latent_function_inference/dtc.py index 5f3e9ff7..6149ce88 100644 --- a/GPy/inference/latent_function_inference/dtc.py +++ b/GPy/inference/latent_function_inference/dtc.py @@ -28,8 +28,8 @@ class DTC(LatentFunctionInference): num_data, output_dim = Y.shape #make sure the noise is not hetero - gaussian_variance = 1./likelihood.gaussian_variance(Y_metadata) - if gaussian_variance.size > 1: + precision = 1./likelihood.gaussian_variance(Y_metadata) + if precision.size > 1: raise NotImplementedError("no hetero noise with this implementation of DTC") Kmm = kern.K(Z) @@ -42,7 +42,7 @@ class DTC(LatentFunctionInference): Kmmi, L, Li, _ = pdinv(Kmm) # Compute A - LiUTbeta = np.dot(Li, U.T)*np.sqrt(gaussian_variance) + LiUTbeta = np.dot(Li, U.T)*np.sqrt(precision) A = tdot(LiUTbeta) + np.eye(num_inducing) # factor A @@ -50,7 +50,7 @@ class DTC(LatentFunctionInference): # back substutue to get b, P, v tmp, _ = dtrtrs(L, Uy, lower=1) - b, _ = dtrtrs(LA, tmp*gaussian_variance, lower=1) + b, _ = dtrtrs(LA, tmp*precision, lower=1) tmp, _ = dtrtrs(LA, b, lower=1, trans=1) v, _ = dtrtrs(L, tmp, lower=1, trans=1) tmp, _ = dtrtrs(LA, Li, lower=1, trans=0) @@ -59,8 +59,8 @@ class DTC(LatentFunctionInference): #compute log marginal log_marginal = -0.5*num_data*output_dim*np.log(2*np.pi) + \ -np.sum(np.log(np.diag(LA)))*output_dim + \ - 0.5*num_data*output_dim*np.log(gaussian_variance) + \ - -0.5*gaussian_variance*np.sum(np.square(Y)) + \ + 0.5*num_data*output_dim*np.log(precision) + \ + -0.5*precision*np.sum(np.square(Y)) + \ 0.5*np.sum(np.square(b)) # Compute dL_dKmm @@ -70,11 +70,11 @@ class DTC(LatentFunctionInference): # Compute dL_dU vY = np.dot(v.reshape(-1,1),Y.T) dL_dU = vY - np.dot(vvT_P, U.T) - dL_dU *= gaussian_variance + dL_dU *= precision #compute dL_dR Uv = np.dot(U, v) - dL_dR = 0.5*(np.sum(U*np.dot(U,P), 1) - 1./gaussian_variance + np.sum(np.square(Y), 1) - 2.*np.sum(Uv*Y, 1) + np.sum(np.square(Uv), 1))*gaussian_variance**2 + dL_dR = 0.5*(np.sum(U*np.dot(U,P), 1) - 1./precision + np.sum(np.square(Y), 1) - 2.*np.sum(Uv*Y, 1) + np.sum(np.square(Uv), 1))*precision**2 dL_dthetaL = likelihood.exact_inference_gradients(dL_dR) @@ -97,8 +97,8 @@ class vDTC(object): num_data, output_dim = Y.shape #make sure the noise is not hetero - gaussian_variance = 1./likelihood.gaussian_variance(Y_metadata) - if gaussian_variance.size > 1: + precision = 1./likelihood.gaussian_variance(Y_metadata) + if precision.size > 1: raise NotImplementedError("no hetero noise with this implementation of DTC") Kmm = kern.K(Z) @@ -111,9 +111,9 @@ class vDTC(object): Kmmi, L, Li, _ = pdinv(Kmm) # Compute A - LiUTbeta = np.dot(Li, U.T)*np.sqrt(gaussian_variance) + LiUTbeta = np.dot(Li, U.T)*np.sqrt(precision) A_ = tdot(LiUTbeta) - trace_term = -0.5*(np.sum(Knn)*gaussian_variance - np.trace(A_)) + trace_term = -0.5*(np.sum(Knn)*precision - np.trace(A_)) A = A_ + np.eye(num_inducing) # factor A @@ -121,7 +121,7 @@ class vDTC(object): # back substutue to get b, P, v tmp, _ = dtrtrs(L, Uy, lower=1) - b, _ = dtrtrs(LA, tmp*gaussian_variance, lower=1) + b, _ = dtrtrs(LA, tmp*precision, lower=1) tmp, _ = dtrtrs(LA, b, lower=1, trans=1) v, _ = dtrtrs(L, tmp, lower=1, trans=1) tmp, _ = dtrtrs(LA, Li, lower=1, trans=0) @@ -131,8 +131,8 @@ class vDTC(object): #compute log marginal log_marginal = -0.5*num_data*output_dim*np.log(2*np.pi) + \ -np.sum(np.log(np.diag(LA)))*output_dim + \ - 0.5*num_data*output_dim*np.log(gaussian_variance) + \ - -0.5*gaussian_variance*np.sum(np.square(Y)) + \ + 0.5*num_data*output_dim*np.log(precision) + \ + -0.5*precision*np.sum(np.square(Y)) + \ 0.5*np.sum(np.square(b)) + \ trace_term @@ -145,15 +145,15 @@ class vDTC(object): vY = np.dot(v.reshape(-1,1),Y.T) #dL_dU = vY - np.dot(vvT_P, U.T) dL_dU = vY - np.dot(vvT_P - Kmmi, U.T) - dL_dU *= gaussian_variance + dL_dU *= precision #compute dL_dR Uv = np.dot(U, v) - dL_dR = 0.5*(np.sum(U*np.dot(U,P), 1) - 1./gaussian_variance + np.sum(np.square(Y), 1) - 2.*np.sum(Uv*Y, 1) + np.sum(np.square(Uv), 1) )*gaussian_variance**2 - dL_dR -=gaussian_variance*trace_term/num_data + dL_dR = 0.5*(np.sum(U*np.dot(U,P), 1) - 1./precision + np.sum(np.square(Y), 1) - 2.*np.sum(Uv*Y, 1) + np.sum(np.square(Uv), 1) )*precision**2 + dL_dR -=precision*trace_term/num_data dL_dthetaL = likelihood.exact_inference_gradients(dL_dR) - grad_dict = {'dL_dKmm': dL_dK, 'dL_dKdiag':np.zeros_like(Knn) + -0.5*gaussian_variance, 'dL_dKnm':dL_dU.T, 'dL_dthetaL':dL_dthetaL} + grad_dict = {'dL_dKmm': dL_dK, 'dL_dKdiag':np.zeros_like(Knn) + -0.5*precision, 'dL_dKnm':dL_dU.T, 'dL_dthetaL':dL_dthetaL} #construct a posterior object post = Posterior(woodbury_inv=Kmmi-P, woodbury_vector=v, K=Kmm, mean=None, cov=None, K_chol=L) diff --git a/GPy/inference/latent_function_inference/exact_gaussian_inference.py b/GPy/inference/latent_function_inference/exact_gaussian_inference.py index 375e1e70..2d8fb691 100644 --- a/GPy/inference/latent_function_inference/exact_gaussian_inference.py +++ b/GPy/inference/latent_function_inference/exact_gaussian_inference.py @@ -22,7 +22,7 @@ class ExactGaussianInference(LatentFunctionInference): def __init__(self): pass#self._YYTfactor_cache = caching.cache() - def inference(self, kern, X, likelihood, Y, mean_function=None, Y_metadata=None, K=None, gaussian_variance=None): + def inference(self, kern, X, likelihood, Y, mean_function=None, Y_metadata=None, K=None, precision=None): """ Returns a Posterior class containing essential quantities of the posterior """ @@ -32,8 +32,8 @@ class ExactGaussianInference(LatentFunctionInference): else: m = mean_function.f(X) - if gaussian_variance is None: - gaussian_variance = likelihood.gaussian_variance(Y_metadata) + if precision is None: + precision = likelihood.gaussian_variance(Y_metadata) YYT_factor = Y-m @@ -41,7 +41,7 @@ class ExactGaussianInference(LatentFunctionInference): K = kern.K(X) Ky = K.copy() - diag.add(Ky, gaussian_variance+1e-8) + diag.add(Ky, precision+1e-8) Wi, LW, LWi, W_logdet = pdinv(Ky) diff --git a/GPy/inference/latent_function_inference/expectation_propagation.py b/GPy/inference/latent_function_inference/expectation_propagation.py index 089f325e..bb289e12 100644 --- a/GPy/inference/latent_function_inference/expectation_propagation.py +++ b/GPy/inference/latent_function_inference/expectation_propagation.py @@ -35,7 +35,7 @@ class EP(ExactGaussianInference): # TODO: update approximation in the end as well? Maybe even with a switch? pass - def inference(self, kern, X, likelihood, Y, mean_function=None, Y_metadata=None, gaussian_variance=None, K=None): + def inference(self, kern, X, likelihood, Y, mean_function=None, Y_metadata=None, precision=None, K=None): num_data, output_dim = Y.shape assert output_dim ==1, "ep in 1D only (for now!)" @@ -49,7 +49,7 @@ class EP(ExactGaussianInference): #if we've already run EP, just use the existing approximation stored in self._ep_approximation mu, Sigma, mu_tilde, tau_tilde, Z_hat = self._ep_approximation - return super(EP, self).inference(kern, X, likelihood, mu_tilde[:,None], mean_function=mean_function, Y_metadata=Y_metadata, gaussian_variance=1./tau_tilde, K=K) + return super(EP, self).inference(kern, X, likelihood, mu_tilde[:,None], mean_function=mean_function, Y_metadata=Y_metadata, precision=1./tau_tilde, K=K) def expectation_propagation(self, K, Y, likelihood, Y_metadata): diff --git a/GPy/inference/latent_function_inference/expectation_propagation_dtc.py b/GPy/inference/latent_function_inference/expectation_propagation_dtc.py index 7b64d8c5..0f141824 100644 --- a/GPy/inference/latent_function_inference/expectation_propagation_dtc.py +++ b/GPy/inference/latent_function_inference/expectation_propagation_dtc.py @@ -46,7 +46,7 @@ class EPDTC(VarDTC): return super(EPDTC, self).inference(kern, X, Z, likelihood, mu_tilde, mean_function=mean_function, Y_metadata=Y_metadata, - gaussian_variance=tau_tilde, + precision=tau_tilde, Lm=Lm, dL_dKmm=dL_dKmm, psi0=psi0, psi1=psi1, psi2=psi2) diff --git a/GPy/inference/latent_function_inference/var_dtc.py b/GPy/inference/latent_function_inference/var_dtc.py index c50f0e7d..bb114050 100644 --- a/GPy/inference/latent_function_inference/var_dtc.py +++ b/GPy/inference/latent_function_inference/var_dtc.py @@ -64,7 +64,7 @@ class VarDTC(LatentFunctionInference): def get_VVTfactor(self, Y, prec): return Y * prec # TODO chache this, and make it effective - def inference(self, kern, X, Z, likelihood, Y, Y_metadata=None, mean_function=None, gaussian_variance=None, Lm=None, dL_dKmm=None, psi0=None, psi1=None, psi2=None): + def inference(self, kern, X, Z, likelihood, Y, Y_metadata=None, mean_function=None, precision=None, Lm=None, dL_dKmm=None, psi0=None, psi1=None, psi2=None): assert mean_function is None, "inference with a mean function not implemented" num_data, output_dim = Y.shape @@ -72,16 +72,16 @@ class VarDTC(LatentFunctionInference): uncertain_inputs = isinstance(X, VariationalPosterior) - if gaussian_variance is None: + if precision is None: #assume Gaussian likelihood - gaussian_variance = 1./np.fmax(likelihood.gaussian_variance(Y_metadata), self.const_jitter) + precision = 1./np.fmax(likelihood.gaussian_variance(Y_metadata), self.const_jitter) - if gaussian_variance.ndim == 1: - gaussian_variance = gaussian_variance[:, None] - het_noise = gaussian_variance.size > 1 + if precision.ndim == 1: + precision = precision[:, None] + het_noise = precision.size > 1 - VVT_factor = gaussian_variance*Y - #VVT_factor = gaussian_variance*Y + VVT_factor = precision*Y + #VVT_factor = precision*Y trYYT = self.get_trYYT(Y) # kernel computations, using BGPLVM notation @@ -98,16 +98,16 @@ class VarDTC(LatentFunctionInference): psi1 = kern.psi1(Z, X) if het_noise: if psi2 is None: - psi2_beta = (kern.psi2n(Z, X) * gaussian_variance[:, :, None]).sum(0) + psi2_beta = (kern.psi2n(Z, X) * precision[:, :, None]).sum(0) else: - psi2_beta = (psi2 * gaussian_variance[:, :, None]).sum(0) + psi2_beta = (psi2 * precision[:, :, None]).sum(0) else: if psi2 is None: - psi2_beta = kern.psi2(Z,X) * gaussian_variance + psi2_beta = kern.psi2(Z,X) * precision elif psi2.ndim == 3: - psi2_beta = psi2.sum(0) * gaussian_variance + psi2_beta = psi2.sum(0) * precision else: - psi2_beta = psi2 * gaussian_variance + psi2_beta = psi2 * precision LmInv = dtrtri(Lm) A = LmInv.dot(psi2_beta.dot(LmInv.T)) else: @@ -116,9 +116,9 @@ class VarDTC(LatentFunctionInference): if psi1 is None: psi1 = kern.K(X, Z) if het_noise: - tmp = psi1 * (np.sqrt(gaussian_variance)) + tmp = psi1 * (np.sqrt(precision)) else: - tmp = psi1 * (np.sqrt(gaussian_variance)) + tmp = psi1 * (np.sqrt(precision)) tmp, _ = dtrtrs(Lm, tmp.T, lower=1) A = tdot(tmp) #print A.sum() @@ -144,19 +144,19 @@ class VarDTC(LatentFunctionInference): dL_dKmm = backsub_both_sides(Lm, delit) # derivatives of L w.r.t. psi - dL_dpsi0, dL_dpsi1, dL_dpsi2 = _compute_dL_dpsi(num_inducing, num_data, output_dim, gaussian_variance, Lm, + dL_dpsi0, dL_dpsi1, dL_dpsi2 = _compute_dL_dpsi(num_inducing, num_data, output_dim, precision, Lm, VVT_factor, Cpsi1Vf, DBi_plus_BiPBi, psi1, het_noise, uncertain_inputs) # log marginal likelihood - log_marginal = _compute_log_marginal_likelihood(likelihood, num_data, output_dim, gaussian_variance, het_noise, + log_marginal = _compute_log_marginal_likelihood(likelihood, num_data, output_dim, precision, het_noise, psi0, A, LB, trYYT, data_fit, Y) #noise derivatives dL_dR = _compute_dL_dR(likelihood, het_noise, uncertain_inputs, LB, _LBi_Lmi_psi1Vf, DBi_plus_BiPBi, Lm, A, - psi0, psi1, gaussian_variance, + psi0, psi1, precision, data_fit, num_data, output_dim, trYYT, Y, VVT_factor) dL_dthetaL = likelihood.exact_inference_gradients(dL_dR,Y_metadata) @@ -181,7 +181,7 @@ class VarDTC(LatentFunctionInference): else: print('foobar') import ipdb; ipdb.set_trace() - psi1V = np.dot(Y.T*gaussian_variance, psi1).T + psi1V = np.dot(Y.T*precision, psi1).T tmp, _ = dtrtrs(Lm, psi1V, lower=1, trans=0) tmp, _ = dpotrs(LB, tmp, lower=1) woodbury_vector, _ = dtrtrs(Lm, tmp, lower=1, trans=1)