apunkt/GPy
1
0
Fork 0
mirror of https://github.com/SheffieldML/GPy.git synced 2026-05-08 11:32:39 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

update additive kernel for SSGPLVM

Browse source
This commit is contained in:
Zhenwen Dai 2014-08-18 18:14:16 +01:00
parent f5f90b3a2c
commit b9fdbedf20
1 changed files with 10 additions and 12 deletions
Download patch file Download diff file Expand all files Collapse all files

22
GPy/kern/_src/add.py
View file

@ -118,9 +118,9 @@ class Add(CombinationKernel):
if isinstance(p2, White): if isinstance(p2, White):
continue continue
elif isinstance(p2, Bias): elif isinstance(p2, Bias):
eff_dL_dpsi1 += dL_dpsi2.sum(1) * p2.variance * 2. eff_dL_dpsi1 += dL_dpsi2.sum(0) * p2.variance * 2.
else:# np.setdiff1d(p1.active_dims, ar2, assume_unique): # TODO: Careful, not correct for overlapping active_dims else:# np.setdiff1d(p1.active_dims, ar2, assume_unique): # TODO: Careful, not correct for overlapping active_dims
eff_dL_dpsi1 += dL_dpsi2.sum(1) * p2.psi1(Z, variational_posterior) * 2. eff_dL_dpsi1 += dL_dpsi2.sum(0) * p2.psi1(Z, variational_posterior) * 2.
p1.update_gradients_expectations(dL_dpsi0, eff_dL_dpsi1, dL_dpsi2, Z, variational_posterior) p1.update_gradients_expectations(dL_dpsi0, eff_dL_dpsi1, dL_dpsi2, Z, variational_posterior)
def gradients_Z_expectations(self, dL_psi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior): def gradients_Z_expectations(self, dL_psi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
@ -135,16 +135,15 @@ class Add(CombinationKernel):
if isinstance(p2, White): if isinstance(p2, White):
continue continue
elif isinstance(p2, Bias): elif isinstance(p2, Bias):
eff_dL_dpsi1 += dL_dpsi2.sum(1) * p2.variance * 2. eff_dL_dpsi1 += dL_dpsi2.sum(0) * p2.variance * 2.
else: else:
eff_dL_dpsi1 += dL_dpsi2.sum(1) * p2.psi1(Z, variational_posterior) * 2. eff_dL_dpsi1 += dL_dpsi2.sum(0) * p2.psi1(Z, variational_posterior) * 2.
target += p1.gradients_Z_expectations(dL_psi0, eff_dL_dpsi1, dL_dpsi2, Z, variational_posterior) target += p1.gradients_Z_expectations(dL_psi0, eff_dL_dpsi1, dL_dpsi2, Z, variational_posterior)
return target return target
def gradients_qX_expectations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior): def gradients_qX_expectations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
from static import White, Bias from static import White, Bias
target_mu = np.zeros(variational_posterior.shape) target_grads = [np.zeros(v.shape) for v in variational_posterior.parameters]
target_S = np.zeros(variational_posterior.shape)
for p1 in self.parameters: for p1 in self.parameters:
#compute the effective dL_dpsi1. extra terms appear becaue of the cross terms in psi2! #compute the effective dL_dpsi1. extra terms appear becaue of the cross terms in psi2!
eff_dL_dpsi1 = dL_dpsi1.copy() eff_dL_dpsi1 = dL_dpsi1.copy()
@ -154,13 +153,12 @@ class Add(CombinationKernel):
if isinstance(p2, White): if isinstance(p2, White):
continue continue
elif isinstance(p2, Bias): elif isinstance(p2, Bias):
eff_dL_dpsi1 += dL_dpsi2.sum(1) * p2.variance * 2. eff_dL_dpsi1 += dL_dpsi2.sum(0) * p2.variance * 2.
else: else:
eff_dL_dpsi1 += dL_dpsi2.sum(1) * p2.psi1(Z, variational_posterior) * 2. eff_dL_dpsi1 += dL_dpsi2.sum(0) * p2.psi1(Z, variational_posterior) * 2.
a, b = p1.gradients_qX_expectations(dL_dpsi0, eff_dL_dpsi1, dL_dpsi2, Z, variational_posterior) grads = p1.gradients_qX_expectations(dL_dpsi0, eff_dL_dpsi1, dL_dpsi2, Z, variational_posterior)
target_mu += a [np.add(target_grads[i],grads[i],target_grads[i]) for i in xrange(len(grads))]
target_S += b return target_grads
return target_mu, target_S
def add(self, other, name='sum'): def add(self, other, name='sum'):
if isinstance(other, Add): if isinstance(other, Add):