apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-14 14:32:37 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

predictino working nicely for laplace

Browse source
This commit is contained in:
James Hensman 2014-02-05 17:52:17 +00:00
parent 80629e00b6
commit 75241ecf89
2 changed files with 15 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

9
GPy/core/gp.py
View file

@ -76,14 +76,17 @@ class GP(Model):
""" """
Kx = self.kern.K(_Xnew, self.X, which_parts=which_parts).T Kx = self.kern.K(_Xnew, self.X, which_parts=which_parts).T
LiKx, _ = dtrtrs(self.posterior.woodbury_chol, np.asfortranarray(Kx), lower=1) #LiKx, _ = dtrtrs(self.posterior.woodbury_chol, np.asfortranarray(Kx), lower=1)
WiKx = np.dot(self.posterior.woodbury_inv, Kx)
mu = np.dot(Kx.T, self.posterior.woodbury_vector) mu = np.dot(Kx.T, self.posterior.woodbury_vector)
if full_cov: if full_cov:
Kxx = self.kern.K(_Xnew, which_parts=which_parts) Kxx = self.kern.K(_Xnew, which_parts=which_parts)
var = Kxx - tdot(LiKx.T) #var = Kxx - tdot(LiKx.T)
var = np.dot(Kx.T, WiKx)
else: else:
Kxx = self.kern.Kdiag(_Xnew, which_parts=which_parts) Kxx = self.kern.Kdiag(_Xnew, which_parts=which_parts)
var = Kxx - np.sum(LiKx*LiKx, 0) #var = Kxx - np.sum(LiKx*LiKx, 0)
var = Kxx - np.sum(WiKx*Kx, 0)
var = var.reshape(-1, 1) var = var.reshape(-1, 1)
return mu, var return mu, var

11
GPy/inference/latent_function_inference/laplace.py
View file

@ -11,7 +11,7 @@
#http://gaussianprocess.org/gpml/code. #http://gaussianprocess.org/gpml/code.
import numpy as np import numpy as np
from ...util.linalg import mdot, jitchol, pddet, dpotrs, dtrtrs from ...util.linalg import mdot, jitchol, pddet, dpotrs, dtrtrs, dpotri, symmetrify
from ...util.misc import param_to_array from ...util.misc import param_to_array
from functools import partial as partial_func from functools import partial as partial_func
from posterior import Posterior from posterior import Posterior
@ -216,8 +216,15 @@ class LaplaceInference(object):
B = np.eye(K.shape[0]) + W_12*K*W_12.T B = np.eye(K.shape[0]) + W_12*K*W_12.T
L = jitchol(B) L = jitchol(B)
LiW12, _ = dtrtrs(L, np.diag(W_12[:,0]), lower=1, trans=0) LiW12, _ = dtrtrs(L, np.diagflat(W_12), lower=1, trans=0)
K_Wi_i = np.dot(LiW12.T, LiW12) # R = W12BiW12, in R&W p 126, eq 5.25 K_Wi_i = np.dot(LiW12.T, LiW12) # R = W12BiW12, in R&W p 126, eq 5.25
#here's a better way to compute the required matrix.
# you could do the model finding witha backsub, instead of a dot...
#L2 = L/W_12
#K_Wi_i_2 , _= dpotri(L2)
#symmetrify(K_Wi_i_2)
return K_Wi_i, L, LiW12 return K_Wi_i, L, LiW12