From fdb7b99e0bd8a740dd898317aab5cd506b97e34e Mon Sep 17 00:00:00 2001 From: Alan Saul Date: Mon, 29 Jul 2013 17:21:52 +0100 Subject: [PATCH] Got rid of some overdoing the approximation --- GPy/likelihoods/Laplace.py | 2 +- GPy/models/GP.py | 10 +++++----- 2 files changed, 6 insertions(+), 6 deletions(-) diff --git a/GPy/likelihoods/Laplace.py b/GPy/likelihoods/Laplace.py index 8b39f222..f86c47b6 100644 --- a/GPy/likelihoods/Laplace.py +++ b/GPy/likelihoods/Laplace.py @@ -165,7 +165,7 @@ class Laplace(likelihood): self.aA = 0.5*self.ln_det_K_Wi__Bi self.bB = - 0.5*self.f_Ki_f self.cC = 0.5*self.y_Wi_Ki_i_y - Z_tilde = (+ 100*self.NORMAL_CONST + Z_tilde = (#+ 100*self.NORMAL_CONST + self.lik + 0.5*self.ln_det_K_Wi__Bi - 0.5*self.f_Ki_f diff --git a/GPy/models/GP.py b/GPy/models/GP.py index 0f56e21c..77620488 100644 --- a/GPy/models/GP.py +++ b/GPy/models/GP.py @@ -132,9 +132,9 @@ class GP(model): model for a new variable Y* = v_tilde/tau_tilde, with a covariance matrix K* = K + diag(1./tau_tilde) plus a normalization term. """ - if isinstance(self.likelihood, Laplace): - self.likelihood.fit_full(self.kern.K(self.X)) - self.likelihood._set_params(self.likelihood._get_params()) + #if isinstance(self.likelihood, Laplace): + #self.likelihood.fit_full(self.kern.K(self.X)) + #self.likelihood._set_params(self.likelihood._get_params()) l = -0.5 * self.D * self.K_logdet + self._model_fit_term() + self.likelihood.Z print "K_ldet: {} mft: {} Z: {}".format(self.K_logdet, self._model_fit_term(), self.likelihood.Z) return l @@ -148,8 +148,8 @@ class GP(model): dL_dthetaK = self.kern.dK_dtheta(dL_dK=self.dL_dK, X=self.X) print "dL_dthetaK should be: ", dL_dthetaK if isinstance(self.likelihood, Laplace): - self.likelihood.fit_full(self.kern.K(self.X)) - self.likelihood._set_params(self.likelihood._get_params()) + #self.likelihood.fit_full(self.kern.K(self.X)) + #self.likelihood._set_params(self.likelihood._get_params()) dK_dthetaK = self.kern.dK_dtheta dL_dthetaK = self.likelihood._Kgradients(dK_dthetaK, self.X.copy()) dL_dthetaL = self.likelihood._gradients(partial=np.diag(self.dL_dK))