apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-24 14:15:14 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

fixed bug in sparse GP plotting

Browse source
This commit is contained in:
Nicolo Fusi 2013-05-21 16:36:12 +01:00
parent b59253fe01
commit 17a82f99bd
3 changed files with 2 additions and 33 deletions
Download patch file Download diff file Expand all files Collapse all files

31
GPy/inference/SGD.py
View file

@ -257,37 +257,6 @@ class opt_SGD(Optimizer):
self.learning_rate = np.ones_like(self.learning_rate)*(np.dot(self.gbar_t.T, self.gbar_t) / self.hbar_t)
tau_t = self.tau_t*(1-self.learning_rate) + 1
# if t == 0:
# N = self.model.N
# Q = self.model.Q
# M = self.model.M
# iip_pos = np.arange(2*N*Q,2*N*Q+M*Q)
# mu_pos = np.arange(0,N*Q)
# S_pos = np.arange(N*Q,2*N*Q)
# self.vbparam_dict = {'iip': [iip_pos],
# 'mu': [mu_pos],
# 'S': [S_pos]}
# for k in self.vbparam_dict.keys():
# hbar_t = 0.0
# tau_t = 1.0
# gbar_t = 0.0
# self.vbparam_dict[k].append(hbar_t)
# self.vbparam_dict[k].append(tau_t)
# self.vbparam_dict[k].append(gbar_t)
# if True:
# g_t = self.model.grads
# for k in self.vbparam_dict.keys():
# pos, hbar_t, tau_t, gbar_t = self.vbparam_dict[k]
# gbar_t = (1-1/tau_t)*gbar_t + 1/tau_t * g_t[pos]
# hbar_t = (1-1/tau_t)*hbar_t + 1/tau_t * np.dot(g_t[pos].T, g_t[pos])
# self.learning_rate[pos] = (np.dot(gbar_t.T, gbar_t) / hbar_t)*1.0
# tau_t = tau_t*(1-self.learning_rate[pos]) + 1
# self.vbparam_dict[k] = [pos, hbar_t, tau_t, gbar_t]
# print k, self.learning_rate[pos].max()
def opt(self, f_fp=None, f=None, fp=None):
self.x_opt = self.model._get_params_transformed()

2
GPy/models/GP.py
View file

@ -173,7 +173,7 @@ class GP(model):
"""
# normalize X values
Xnew = (Xnew.copy() - self._Xmean) / self._Xstd
mu, var = self._raw_predict(Xnew, which_parts, full_cov)
mu, var = self._raw_predict(Xnew, which_parts=which_parts, full_cov=full_cov)
# now push through likelihood
mean, var, _025pm, _975pm = self.likelihood.predictive_values(mu, var, full_cov)

2
GPy/models/sparse_GP.py
View file

@ -234,7 +234,7 @@ class sparse_GP(GP):
Kxx = self.kern.Kdiag(Xnew, which_parts=which_parts)
var = Kxx - np.sum(Kx * np.dot(Kmmi_LmiBLmi, Kx), 0)
else:
assert which_parts=='all', "swithching out parts of variational kernels is not implemented"
# assert which_parts=='all', "swithching out parts of variational kernels is not implemented"
Kx = self.kern.psi1(self.Z, Xnew, X_variance_new)#, which_parts=which_parts) TODO: which_parts
mu = np.dot(Kx, self.Cpsi1V)
if full_cov: