mirror of
https://github.com/SheffieldML/GPy.git
synced 2026-06-11 15:15:15 +02:00
fall back to the old psicomp interfacegit status!
This commit is contained in:
Zhenwen Dai
parent
018c20400d
commit
901628a30c
6 changed files with 37 additions and 70 deletions
|
|
@ -103,8 +103,7 @@ class Kern(Parameterized):
|
|||
"""Set the gradients of all parameters when doing full (N) inference."""
|
||||
raise NotImplementedError
|
||||
|
||||
def update_gradients_expectations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior,
|
||||
psi0=None, psi1=None, psi2=None):
|
||||
def update_gradients_expectations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
|
||||
"""
|
||||
Set the gradients of all parameters when doing inference with
|
||||
uncertain inputs, using expectations of the kernel.
|
||||
|
|
@ -115,8 +114,7 @@ class Kern(Parameterized):
|
|||
dL_dpsi1 * dpsi1_d{theta_i} +
|
||||
dL_dpsi2 * dpsi2_d{theta_i}
|
||||
"""
|
||||
dtheta = self.psicomp.psiDerivativecomputations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior,
|
||||
psi0=psi0, psi1=psi1, psi2=psi2)[0]
|
||||
dtheta = self.psicomp.psiDerivativecomputations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior)[0]
|
||||
self.gradient[:] = dtheta
|
||||
|
||||
def gradients_Z_expectations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior,
|
||||
|
|
@ -125,17 +123,14 @@ class Kern(Parameterized):
|
|||
Returns the derivative of the objective wrt Z, using the chain rule
|
||||
through the expectation variables.
|
||||
"""
|
||||
return self.psicomp.psiDerivativecomputations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior,
|
||||
psi0=psi0, psi1=psi1, psi2=psi2)[1]
|
||||
return self.psicomp.psiDerivativecomputations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior)[1]
|
||||
|
||||
def gradients_qX_expectations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior,
|
||||
psi0=None, psi1=None, psi2=None, Lpsi0=None, Lpsi1=None, Lpsi2=None):
|
||||
def gradients_qX_expectations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
|
||||
"""
|
||||
Compute the gradients wrt the parameters of the variational
|
||||
distruibution q(X), chain-ruling via the expectations of the kernel
|
||||
"""
|
||||
return self.psicomp.psiDerivativecomputations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior,
|
||||
psi0=psi0, psi1=psi1, psi2=psi2)[2:]
|
||||
return self.psicomp.psiDerivativecomputations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior)[2:]
|
||||
|
||||
def plot(self, x=None, fignum=None, ax=None, title=None, plot_limits=None, resolution=None, **mpl_kwargs):
|
||||
"""
|
||||
|
|
|
|||
|
|
@ -137,11 +137,9 @@ def _slice_psi(f):
|
|||
|
||||
def _slice_update_gradients_expectations(f):
|
||||
@wraps(f)
|
||||
def wrap(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior,
|
||||
psi0=None, psi1=None, psi2=None, Lpsi0=None, Lpsi1=None, Lpsi2=None):
|
||||
def wrap(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
|
||||
with _Slice_wrap(self, Z, variational_posterior) as s:
|
||||
ret = f(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, s.X, s.X2,
|
||||
psi0=psi0, psi1=psi1, psi2=psi2, Lpsi0=Lpsi0, Lpsi1=Lpsi1, Lpsi2=Lpsi2)
|
||||
ret = f(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, s.X, s.X2)
|
||||
return ret
|
||||
return wrap
|
||||
|
||||
|
|
@ -150,8 +148,7 @@ def _slice_gradients_Z_expectations(f):
|
|||
def wrap(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior,
|
||||
psi0=None, psi1=None, psi2=None, Lpsi0=None, Lpsi1=None, Lpsi2=None):
|
||||
with _Slice_wrap(self, Z, variational_posterior) as s:
|
||||
ret = s.handle_return_array(f(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, s.X, s.X2,
|
||||
psi0=psi0, psi1=psi1, psi2=psi2, Lpsi0=Lpsi0, Lpsi1=Lpsi1, Lpsi2=Lpsi2))
|
||||
ret = s.handle_return_array(f(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, s.X, s.X2))
|
||||
return ret
|
||||
return wrap
|
||||
|
||||
|
|
@ -160,8 +157,7 @@ def _slice_gradients_qX_expectations(f):
|
|||
def wrap(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior,
|
||||
psi0=None, psi1=None, psi2=None, Lpsi0=None, Lpsi1=None, Lpsi2=None):
|
||||
with _Slice_wrap(self, variational_posterior, Z) as s:
|
||||
ret = list(f(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, s.X2, s.X,
|
||||
psi0=psi0, psi1=psi1, psi2=psi2, Lpsi0=Lpsi0, Lpsi1=Lpsi1, Lpsi2=Lpsi2))
|
||||
ret = list(f(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, s.X2, s.X))
|
||||
r2 = ret[:2]
|
||||
ret[0] = s.handle_return_array(r2[0])
|
||||
ret[1] = s.handle_return_array(r2[1])
|
||||
|
|
|
|||
|
|
@ -11,23 +11,19 @@ from . import linear_psi_comp
|
|||
from .gaussherm import PSICOMP_GH
|
||||
|
||||
class PSICOMP_RBF(Pickleable):
|
||||
@Cache_this(limit=10, ignore_args=(0,))
|
||||
@Cache_this(limit=2, ignore_args=(0,))
|
||||
def psicomputations(self, variance, lengthscale, Z, variational_posterior, return_psi2_n=False):
|
||||
if isinstance(variational_posterior, variational.NormalPosterior):
|
||||
return rbf_psi_comp.psicomputations(variance, lengthscale, Z, variational_posterior, return_psi2_n=return_psi2_n)
|
||||
elif isinstance(variational_posterior, variational.SpikeAndSlabPosterior):
|
||||
if return_psi2_n:
|
||||
raise NotImplementedError('However this function seems to return it by default')
|
||||
return ssrbf_psi_comp.psicomputations(variance, lengthscale, Z, variational_posterior)
|
||||
else:
|
||||
raise ValueError("unknown distriubtion received for psi-statistics")
|
||||
|
||||
@Cache_this(limit=10, ignore_args=(0,1,2,3))
|
||||
def psiDerivativecomputations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, variance, lengthscale, Z, variational_posterior,
|
||||
psi0=None, psi1=None, psi2=None, Lpsi0=None, Lpsi1=None, Lpsi2=None):
|
||||
@Cache_this(limit=2, ignore_args=(0,1,2,3))
|
||||
def psiDerivativecomputations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, variance, lengthscale, Z, variational_posterior):
|
||||
if isinstance(variational_posterior, variational.NormalPosterior):
|
||||
return rbf_psi_comp.psiDerivativecomputations(dL_dpsi0, dL_dpsi1, dL_dpsi2, variance, lengthscale, Z, variational_posterior,
|
||||
psi0=psi0, psi1=psi1, psi2=psi2, Lpsi0=Lpsi0, Lpsi1=Lpsi1, Lpsi2=Lpsi2)
|
||||
return rbf_psi_comp.psiDerivativecomputations(dL_dpsi0, dL_dpsi1, dL_dpsi2, variance, lengthscale, Z, variational_posterior)
|
||||
elif isinstance(variational_posterior, variational.SpikeAndSlabPosterior):
|
||||
return ssrbf_psi_comp.psiDerivativecomputations(dL_dpsi0, dL_dpsi1, dL_dpsi2, variance, lengthscale, Z, variational_posterior)
|
||||
else:
|
||||
|
|
@ -38,10 +34,8 @@ class PSICOMP_RBF(Pickleable):
|
|||
|
||||
class PSICOMP_Linear(Pickleable):
|
||||
|
||||
@Cache_this(limit=10, ignore_args=(0,))
|
||||
def psicomputations(self, variance, Z, variational_posterior, return_psi2_n=False):
|
||||
if return_psi2_n:
|
||||
raise NotImplementedError
|
||||
@Cache_this(limit=2, ignore_args=(0,))
|
||||
def psicomputations(self, variance, Z, variational_posterior):
|
||||
if isinstance(variational_posterior, variational.NormalPosterior):
|
||||
return linear_psi_comp.psicomputations(variance, Z, variational_posterior)
|
||||
elif isinstance(variational_posterior, variational.SpikeAndSlabPosterior):
|
||||
|
|
@ -49,10 +43,9 @@ class PSICOMP_Linear(Pickleable):
|
|||
else:
|
||||
raise ValueError("unknown distriubtion received for psi-statistics")
|
||||
|
||||
@Cache_this(limit=10, ignore_args=(0,1,2,3))
|
||||
def psiDerivativecomputations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, variance, Z, variational_posterior, psi0=None, psi1=None, psi2=None):
|
||||
@Cache_this(limit=2, ignore_args=(0,1,2,3))
|
||||
def psiDerivativecomputations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, variance, Z, variational_posterior):
|
||||
if isinstance(variational_posterior, variational.NormalPosterior):
|
||||
#Should pass psi in
|
||||
return linear_psi_comp.psiDerivativecomputations(dL_dpsi0, dL_dpsi1, dL_dpsi2, variance, Z, variational_posterior)
|
||||
elif isinstance(variational_posterior, variational.SpikeAndSlabPosterior):
|
||||
return sslinear_psi_comp.psiDerivativecomputations(dL_dpsi0, dL_dpsi1, dL_dpsi2, variance, Z, variational_posterior)
|
||||
|
|
@ -60,4 +53,4 @@ class PSICOMP_Linear(Pickleable):
|
|||
raise ValueError("unknown distriubtion received for psi-statistics")
|
||||
|
||||
def _setup_observers(self):
|
||||
pass
|
||||
pass
|
||||
|
|
@ -22,7 +22,7 @@ def psicomputations(variance, Z, variational_posterior, return_psi2_n=False):
|
|||
|
||||
psi0 = (variance*(np.square(mu)+S)).sum(axis=1)
|
||||
psi1 = np.dot(mu,(variance*Z).T)
|
||||
if return_psi2_n:
|
||||
if not return_psi2_n:
|
||||
psi2 = np.dot(S.sum(axis=0)*np.square(variance)*Z,Z.T)+ tdot(psi1.T)
|
||||
else:
|
||||
raise NotImplementedError
|
||||
|
|
|
|||
|
|
@ -22,8 +22,7 @@ def psicomputations(variance, lengthscale, Z, variational_posterior, return_psi2
|
|||
psi0[:] = variance
|
||||
psi1 = _psi1computations(variance, lengthscale, Z, mu, S)
|
||||
psi2 = _psi2computations(variance, lengthscale, Z, mu, S)
|
||||
if not return_psi2_n:
|
||||
psi2 = psi2.sum(axis=0)
|
||||
if not return_psi2_n: psi2 = psi2.sum(axis=0)
|
||||
return psi0, psi1, psi2
|
||||
|
||||
def __psi1computations(variance, lengthscale, Z, mu, S):
|
||||
|
|
@ -68,12 +67,11 @@ def __psi2computations(variance, lengthscale, Z, mu, S):
|
|||
_psi2 = variance*variance*np.exp(_psi2_logdenom[:,None,None]+_psi2_exp1[None,:,:]+_psi2_exp2)
|
||||
return _psi2
|
||||
|
||||
def psiDerivativecomputations(dL_dpsi0, dL_dpsi1, dL_dpsi2, variance, lengthscale, Z, variational_posterior,
|
||||
psi0=None, psi1=None, psi2=None, Lpsi0=None, Lpsi1=None, Lpsi2=None):
|
||||
def psiDerivativecomputations(dL_dpsi0, dL_dpsi1, dL_dpsi2, variance, lengthscale, Z, variational_posterior):
|
||||
ARD = (len(lengthscale)!=1)
|
||||
|
||||
dvar_psi1, dl_psi1, dZ_psi1, dmu_psi1, dS_psi1 = _psi1compDer(dL_dpsi1, variance, lengthscale, Z, variational_posterior.mean, variational_posterior.variance, psi1=psi1, Lpsi1=Lpsi1)
|
||||
dvar_psi2, dl_psi2, dZ_psi2, dmu_psi2, dS_psi2 = _psi2compDer(dL_dpsi2, variance, lengthscale, Z, variational_posterior.mean, variational_posterior.variance, psi2=psi2, Lpsi2=Lpsi2)
|
||||
dvar_psi1, dl_psi1, dZ_psi1, dmu_psi1, dS_psi1 = _psi1compDer(dL_dpsi1, variance, lengthscale, Z, variational_posterior.mean, variational_posterior.variance)
|
||||
dvar_psi2, dl_psi2, dZ_psi2, dmu_psi2, dS_psi2 = _psi2compDer(dL_dpsi2, variance, lengthscale, Z, variational_posterior.mean, variational_posterior.variance)
|
||||
|
||||
dL_dvar = np.sum(dL_dpsi0) + dvar_psi1 + dvar_psi2
|
||||
|
||||
|
|
@ -87,7 +85,7 @@ def psiDerivativecomputations(dL_dpsi0, dL_dpsi1, dL_dpsi2, variance, lengthscal
|
|||
|
||||
return dL_dvar, dL_dlengscale, dL_dZ, dL_dmu, dL_dS
|
||||
|
||||
def __psi1compDer(dL_dpsi1, variance, lengthscale, Z, mu, S, psi1=None, Lpsi1=None):
|
||||
def _psi1compDer(dL_dpsi1, variance, lengthscale, Z, mu, S):
|
||||
"""
|
||||
dL_dpsi1 - NxM
|
||||
Z - MxQ
|
||||
|
|
@ -106,10 +104,8 @@ def __psi1compDer(dL_dpsi1, variance, lengthscale, Z, mu, S, psi1=None, Lpsi1=No
|
|||
|
||||
lengthscale2 = np.square(lengthscale)
|
||||
|
||||
if psi1 is None:
|
||||
psi1 = _psi1computations(variance, lengthscale, Z, mu, S)
|
||||
if Lpsi1 is None:
|
||||
Lpsi1 = dL_dpsi1*psi1
|
||||
_psi1 = _psi1computations(variance, lengthscale, Z, mu, S)
|
||||
Lpsi1 = dL_dpsi1*_psi1
|
||||
Zmu = Z[None,:,:]-mu[:,None,:] # NxMxQ
|
||||
denom = 1./(S+lengthscale2)
|
||||
Zmu2_denom = np.square(Zmu)*denom[:,None,:] #NxMxQ
|
||||
|
|
@ -121,7 +117,7 @@ def __psi1compDer(dL_dpsi1, variance, lengthscale, Z, mu, S, psi1=None, Lpsi1=No
|
|||
|
||||
return _dL_dvar, _dL_dl, _dL_dZ, _dL_dmu, _dL_dS
|
||||
|
||||
def __psi2compDer(dL_dpsi2, variance, lengthscale, Z, mu, S, psi2=None, Lpsi2=None):
|
||||
def _psi2compDer(dL_dpsi2, variance, lengthscale, Z, mu, S):
|
||||
"""
|
||||
Z - MxQ
|
||||
mu - NxQ
|
||||
|
|
@ -142,10 +138,8 @@ def __psi2compDer(dL_dpsi2, variance, lengthscale, Z, mu, S, psi2=None, Lpsi2=No
|
|||
denom = 1./(2*S+lengthscale2)
|
||||
denom2 = np.square(denom)
|
||||
|
||||
if psi2 is None:
|
||||
psi2 = _psi2computations(variance, lengthscale, Z, mu, S) # NxMxM
|
||||
if Lpsi2 is None:
|
||||
Lpsi2 = dL_dpsi2*psi2 # dL_dpsi2 is MxM, using broadcast to multiply N out
|
||||
_psi2 = _psi2computations(variance, lengthscale, Z, mu, S) # NxMxM
|
||||
Lpsi2 = dL_dpsi2*_psi2 # dL_dpsi2 is MxM, using broadcast to multiply N out
|
||||
Lpsi2sum = np.einsum('nmo->n',Lpsi2) #N
|
||||
Lpsi2Z = np.einsum('nmo,oq->nq',Lpsi2,Z) #NxQ
|
||||
Lpsi2Z2 = np.einsum('nmo,oq,oq->nq',Lpsi2,Z,Z) #NxQ
|
||||
|
|
@ -156,14 +150,8 @@ def __psi2compDer(dL_dpsi2, variance, lengthscale, Z, mu, S, psi2=None, Lpsi2=No
|
|||
_dL_dvar = Lpsi2sum.sum()*2/variance
|
||||
_dL_dmu = (-2*denom) * (mu*Lpsi2sum[:,None]-Lpsi2Zhat)
|
||||
_dL_dS = (2*np.square(denom))*(np.square(mu)*Lpsi2sum[:,None]-2*mu*Lpsi2Zhat+Lpsi2Zhat2) - denom*Lpsi2sum[:,None]
|
||||
_dL_dZ1 = -np.einsum('nmo,oq->oq',Lpsi2,Z)/lengthscale2
|
||||
_dL_dZ2 = np.einsum('nmo,oq->mq',Lpsi2,Z)/lengthscale2
|
||||
_dL_dZ3 = 2*np.einsum('nmo,nq,nq->mq',Lpsi2,mu,denom)
|
||||
_dL_dZ4 = - np.einsum('nmo,nq,mq->mq',Lpsi2,denom,Z)
|
||||
_dL_dZ5 = - np.einsum('nmo,oq,nq->mq',Lpsi2,Z,denom)
|
||||
_dL_dZ = _dL_dZ1 + _dL_dZ2 + _dL_dZ3 + _dL_dZ4 + _dL_dZ5
|
||||
#_dL_dZ = -np.einsum('nmo,oq->oq',Lpsi2,Z)/lengthscale2+np.einsum('nmo,oq->mq',Lpsi2,Z)/lengthscale2+ \
|
||||
#2*np.einsum('nmo,nq,nq->mq',Lpsi2,mu,denom) - np.einsum('nmo,nq,mq->mq',Lpsi2,denom,Z) - np.einsum('nmo,oq,nq->mq',Lpsi2,Z,denom)
|
||||
_dL_dZ = -np.einsum('nmo,oq->oq',Lpsi2,Z)/lengthscale2+np.einsum('nmo,oq->mq',Lpsi2,Z)/lengthscale2+ \
|
||||
2*np.einsum('nmo,nq,nq->mq',Lpsi2,mu,denom) - np.einsum('nmo,nq,mq->mq',Lpsi2,denom,Z) - np.einsum('nmo,oq,nq->mq',Lpsi2,Z,denom)
|
||||
_dL_dl = 2*lengthscale* ((S/lengthscale2*denom+np.square(mu*denom))*Lpsi2sum[:,None]+(Lpsi2Z2-Lpsi2Z2p)/(2*np.square(lengthscale2))-
|
||||
(2*mu*denom2)*Lpsi2Zhat+denom2*Lpsi2Zhat2).sum(axis=0)
|
||||
|
||||
|
|
@ -171,5 +159,3 @@ def __psi2compDer(dL_dpsi2, variance, lengthscale, Z, mu, S, psi2=None, Lpsi2=No
|
|||
|
||||
_psi1computations = Cacher(__psi1computations, limit=5)
|
||||
_psi2computations = Cacher(__psi2computations, limit=5)
|
||||
_psi1compDer = Cacher(__psi1compDer, limit=5)
|
||||
_psi2compDer = Cacher(__psi2compDer, limit=5)
|
||||
|
|
|
|||
|
|
@ -64,17 +64,14 @@ class RBF(Stationary):
|
|||
def psi2n(self, Z, variational_posterior):
|
||||
return self.psicomp.psicomputations(self.variance, self.lengthscale, Z, variational_posterior, return_psi2_n=True)[2]
|
||||
|
||||
def update_gradients_expectations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior,
|
||||
psi0=None, psi1=None, psi2=None, Lpsi0=None, Lpsi1=None, Lpsi2=None):
|
||||
dL_dvar, dL_dlengscale = self.psicomp.psiDerivativecomputations(dL_dpsi0, dL_dpsi1, dL_dpsi2, self.variance, self.lengthscale, Z, variational_posterior, psi0=psi0, psi1=psi1, psi2=psi2, Lpsi0=Lpsi0, Lpsi1=Lpsi1, Lpsi2=Lpsi2)[:2]
|
||||
def update_gradients_expectations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
|
||||
dL_dvar, dL_dlengscale = self.psicomp.psiDerivativecomputations(dL_dpsi0, dL_dpsi1, dL_dpsi2, self.variance, self.lengthscale, Z, variational_posterior)[:2]
|
||||
self.variance.gradient = dL_dvar
|
||||
self.lengthscale.gradient = dL_dlengscale
|
||||
|
||||
def gradients_Z_expectations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior,
|
||||
psi0=None, psi1=None, psi2=None, Lpsi0=None, Lpsi1=None, Lpsi2=None):
|
||||
return self.psicomp.psiDerivativecomputations(dL_dpsi0, dL_dpsi1, dL_dpsi2, self.variance, self.lengthscale, Z, variational_posterior, psi0=psi0, psi1=psi1, psi2=psi2, Lpsi0=Lpsi0, Lpsi1=Lpsi1, Lpsi2=Lpsi2)[2]
|
||||
def gradients_Z_expectations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
|
||||
return self.psicomp.psiDerivativecomputations(dL_dpsi0, dL_dpsi1, dL_dpsi2, self.variance, self.lengthscale, Z, variational_posterior)[2]
|
||||
|
||||
def gradients_qX_expectations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior,
|
||||
psi0=None, psi1=None, psi2=None, Lpsi0=None, Lpsi1=None, Lpsi2=None):
|
||||
return self.psicomp.psiDerivativecomputations(dL_dpsi0, dL_dpsi1, dL_dpsi2, self.variance, self.lengthscale, Z, variational_posterior, psi0=psi0, psi1=psi1, psi2=psi2, Lpsi0=Lpsi0, Lpsi1=Lpsi1, Lpsi2=Lpsi2)[3:]
|
||||
def gradients_qX_expectations(self, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
|
||||
return self.psicomp.psiDerivativecomputations(dL_dpsi0, dL_dpsi1, dL_dpsi2, self.variance, self.lengthscale, Z, variational_posterior)[3:]
|
||||
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue