apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-09 12:02:38 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

merge rbf_inv changes

Browse source
This commit is contained in:
Max Zwiessele 2013-07-18 15:53:50 +01:00
commit cdf7735176
3 changed files with 15 additions and 17 deletions
Download patch file Download diff file Expand all files Collapse all files

4
GPy/core/gp.py
View file

@ -128,7 +128,7 @@ class GP(GPBase):
debug_this # @UndefinedVariable debug_this # @UndefinedVariable
return mu, var return mu, var
def predict(self, Xnew, which_parts='all', full_cov=False): def predict(self, Xnew, which_parts='all', full_cov=False, likelihood_args=dict()):
""" """
Predict the function(s) at the new point(s) Xnew. Predict the function(s) at the new point(s) Xnew.
Arguments Arguments
@ -153,6 +153,6 @@ class GP(GPBase):
mu, var = self._raw_predict(Xnew, full_cov=full_cov, which_parts=which_parts) mu, var = self._raw_predict(Xnew, full_cov=full_cov, which_parts=which_parts)
# now push through likelihood # now push through likelihood
mean, var, _025pm, _975pm = self.likelihood.predictive_values(mu, var, full_cov) mean, var, _025pm, _975pm = self.likelihood.predictive_values(mu, var, full_cov, **likelihood_args)
return mean, var, _025pm, _975pm return mean, var, _025pm, _975pm

3
GPy/kern/parts/rbf.py
View file

@ -165,9 +165,8 @@ class RBF(Kernpart):
def dpsi1_dtheta(self, dL_dpsi1, Z, mu, S, target): def dpsi1_dtheta(self, dL_dpsi1, Z, mu, S, target):
self._psi_computations(Z, mu, S) self._psi_computations(Z, mu, S)
denom_deriv = S[:, None, :] / (self.lengthscale ** 3 + self.lengthscale * S[:, None, :])
d_length = self._psi1[:, :, None] * (self.lengthscale * np.square(self._psi1_dist / (self.lengthscale2 + S[:, None, :])) + denom_deriv)
target[0] += np.sum(dL_dpsi1 * self._psi1 / self.variance) target[0] += np.sum(dL_dpsi1 * self._psi1 / self.variance)
d_length = self._psi1[:,:,None] * ((self._psi1_dist_sq - 1.)/(self.lengthscale*self._psi1_denom) +1./self.lengthscale)
dpsi1_dlength = d_length * dL_dpsi1[:, :, None] dpsi1_dlength = d_length * dL_dpsi1[:, :, None]
if not self.ARD: if not self.ARD:
target[1] += dpsi1_dlength.sum() target[1] += dpsi1_dlength.sum()

25
GPy/kern/parts/rbf_inv.py
View file

@ -159,21 +159,20 @@ class RBFInv(RBF):
def dpsi1_dtheta(self, dL_dpsi1, Z, mu, S, target): def dpsi1_dtheta(self, dL_dpsi1, Z, mu, S, target):
self._psi_computations(Z, mu, S) self._psi_computations(Z, mu, S)
# #d_length = self._psi1[:, :, None] * (-0.5 * ((np.square((self._psi1_dist)/(self.inv_lengthscale * S[:,None,:] + 1))) + ((S[:, None, :])/(self.inv_lengthscale * S[:, None, :] + 1)))) tmp = 1 + S[:,None,:]*self.inv_lengthscale2
tmp = 1 + S[:, None, :] * self.inv_lengthscale2 d_inv_length_old = -self._psi1[:,:,None] * ((self._psi1_dist_sq - 1.)/(self.lengthscale*self._psi1_denom) + self.inv_lengthscale)/self.inv_lengthscale2
# inv_len3 = np.power(self.inv_lengthscale,3) d_inv_length = -self._psi1[:,:,None] * ((self._psi1_dist_sq - 1.)/self._psi1_denom + self.lengthscale)
d_length = -(self._psi1[:, :, None] * ((np.square(self._psi1_dist) * self.inv_lengthscale) / (tmp ** 2) + (S[:, None, :] * self.inv_lengthscale) / (tmp)))
target[0] += np.sum(dL_dpsi1 * self._psi1 / self.variance) target[0] += np.sum(dL_dpsi1 * self._psi1 / self.variance)
dpsi1_dlength = d_length * dL_dpsi1[:, :, None] dpsi1_dlength = d_inv_length * dL_dpsi1[:, :, None]
if not self.ARD: if not self.ARD:
target[1] += dpsi1_dlength.sum() # *(-self.lengthscale2) target[1] += dpsi1_dlength.sum()#*(-self.lengthscale2)
else: else:
target[1:] += dpsi1_dlength.sum(0).sum(0) # *(-self.lengthscale2) target[1:] += dpsi1_dlength.sum(0).sum(0)#*(-self.lengthscale2)
# target[1:] = target[1:]*(-self.lengthscale2) #target[1:] = target[1:]*(-self.lengthscale2)
def dpsi1_dZ(self, dL_dpsi1, Z, mu, S, target): def dpsi1_dZ(self, dL_dpsi1, Z, mu, S, target):
self._psi_computations(Z, mu, S) self._psi_computations(Z, mu, S)
dpsi1_dZ = -self._psi1[:, :, None] * ((self.inv_lengthscale2 * self._psi1_dist) / self._psi1_denom) dpsi1_dZ = -self._psi1[:, :, None] * ((self.inv_lengthscale2*self._psi1_dist)/self._psi1_denom)
target += np.sum(dL_dpsi1[:, :, None] * dpsi1_dZ, 0) target += np.sum(dL_dpsi1[:, :, None] * dpsi1_dZ, 0)
def dpsi1_dmuS(self, dL_dpsi1, Z, mu, S, target_mu, target_S): def dpsi1_dmuS(self, dL_dpsi1, Z, mu, S, target_mu, target_S):
@ -186,15 +185,15 @@ class RBFInv(RBF):
"""Shape N,num_inducing,num_inducing,Ntheta""" """Shape N,num_inducing,num_inducing,Ntheta"""
self._psi_computations(Z, mu, S) self._psi_computations(Z, mu, S)
d_var = 2.*self._psi2 / self.variance d_var = 2.*self._psi2 / self.variance
# d_length = 2.*self._psi2[:, :, :, None] * (self._psi2_Zdist_sq * self._psi2_denom + self._psi2_mudist_sq + S[:, None, None, :] / self.lengthscale2) / (self.lengthscale * self._psi2_denom) #d_length = 2.*self._psi2[:, :, :, None] * (self._psi2_Zdist_sq * self._psi2_denom + self._psi2_mudist_sq + S[:, None, None, :] / self.lengthscale2) / (self.lengthscale * self._psi2_denom)
d_length = -2.*self._psi2[:, :, :, None] * (self._psi2_Zdist_sq * self._psi2_denom + self._psi2_mudist_sq + S[:, None, None, :] * self.inv_lengthscale2) / (self.inv_lengthscale * self._psi2_denom) d_length = -2.*self._psi2[:, :, :, None] * (self._psi2_Zdist_sq * self._psi2_denom + self._psi2_mudist_sq + S[:, None, None, :] * self.inv_lengthscale2) / (self.inv_lengthscale * self._psi2_denom)
target[0] += np.sum(dL_dpsi2 * d_var) target[0] += np.sum(dL_dpsi2 * d_var)
dpsi2_dlength = d_length * dL_dpsi2[:, :, :, None] dpsi2_dlength = d_length * dL_dpsi2[:, :, :, None]
if not self.ARD: if not self.ARD:
target[1] += dpsi2_dlength.sum() # *(-self.lengthscale2) target[1] += dpsi2_dlength.sum()#*(-self.lengthscale2)
else: else:
target[1:] += dpsi2_dlength.sum(0).sum(0).sum(0) # *(-self.lengthscale2) target[1:] += dpsi2_dlength.sum(0).sum(0).sum(0)#*(-self.lengthscale2)
# target[1:] = target[1:]*(-self.lengthscale2) #target[1:] = target[1:]*(-self.lengthscale2)
def dpsi2_dZ(self, dL_dpsi2, Z, mu, S, target): def dpsi2_dZ(self, dL_dpsi2, Z, mu, S, target):
self._psi_computations(Z, mu, S) self._psi_computations(Z, mu, S)