apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-06-29 15:59:41 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix the issue of negative prediction variance of normal GP

Browse source
This commit is contained in:
Zhenwen Dai 2016-01-21 11:22:57 +00:00
parent 5f417565fb
commit 8b279175c5
6 changed files with 84 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

3
GPy/inference/latent_function_inference/exact_gaussian_inference.py
View file

@ -1,14 +1,13 @@
# Copyright (c) 2012-2014, GPy authors (see AUTHORS.txt).
# Licensed under the BSD 3-clause license (see LICENSE.txt)
from .posterior import Posterior
from .posterior import PosteriorExact as Posterior
from ...util.linalg import pdinv, dpotrs, tdot
from ...util import diag
import numpy as np
from . import LatentFunctionInference
log_2_pi = np.log(2*np.pi)
class ExactGaussianInference(LatentFunctionInference):
"""
An object for inference when the likelihood is Gaussian.

34
GPy/inference/latent_function_inference/posterior.py
View file

@ -2,7 +2,7 @@
# Licensed under the BSD 3-clause license (see LICENSE.txt)
import numpy as np
from ...util.linalg import pdinv, dpotrs, dpotri, symmetrify, jitchol
from ...util.linalg import pdinv, dpotrs, dpotri, symmetrify, jitchol, dtrtrs, tdot
class Posterior(object):
"""
@ -187,3 +187,35 @@ class Posterior(object):
if self._K_chol is None:
self._K_chol = jitchol(self._K)
return self._K_chol
class PosteriorExact(Posterior):
def _raw_predict(self, kern, Xnew, pred_var, full_cov=False):
Kx = kern.K(pred_var, Xnew)
mu = np.dot(Kx.T, self.woodbury_vector)
if len(mu.shape)==1:
mu = mu.reshape(-1,1)
if full_cov:
Kxx = kern.K(Xnew)
if self._woodbury_chol.ndim == 2:
tmp = dtrtrs(self._woodbury_chol, Kx)[0]
var = Kxx - tdot(tmp.T)
elif self._woodbury_chol.ndim == 3: # Missing data
var = np.empty((Kxx.shape[0],Kxx.shape[1],self._woodbury_chol.shape[2]))
for i in range(var.shape[2]):
tmp = dtrtrs(self._woodbury_chol[:,:,i], Kx)[0]
var[:, :, i] = (Kxx - tdot(tmp.T))
var = var
else:
Kxx = kern.Kdiag(Xnew)
if self._woodbury_chol.ndim == 2:
tmp = dtrtrs(self._woodbury_chol, Kx)[0]
var = (Kxx - np.square(tmp).sum(0))[:,None]
elif self._woodbury_chol.ndim == 3: # Missing data
var = np.empty((Kxx.shape[0],self._woodbury_chol.shape[2]))
for i in range(var.shape[1]):
tmp = dtrtrs(self._woodbury_chol[:,:,i], Kx)[0]
var[:, i] = (Kxx - np.square(tmp).sum(0))
var = var
return mu, var