def __init__(self,
X,
kernF,
kernY,
Y,
K=2,
alpha=1.,
prior_Z='symmetric',
name='MOHGP'):
N, self.D = Y.shape
self.Y = Y
self.X = X
assert X.shape[0] == self.D, "input data don't match observations"
CollapsedMixture.__init__(self, N, K, prior_Z, alpha, name)
self.kernF = kernF
self.kernY = kernY
self.link_parameters(self.kernF, self.kernY)
#initialize kernels
self.Sf = self.kernF.K(self.X)
self.Sy = self.kernY.K(self.X)
self.Sy_inv, self.Sy_chol, self.Sy_chol_inv, self.Sy_logdet = pdinv(
self.Sy + np.eye(self.D) * 1e-6)
#Computations that can be done outside the optimisation loop
self.YYT = self.Y[:, :, np.newaxis] * self.Y[:, np.newaxis, :]
self.YTY = np.dot(self.Y.T, self.Y)
self.do_computations()
def __init__(self,
X,
Y,
K=2,
kernels=None,
variance=1.,
alpha=1.,
prior_Z='symmetric',
name='OMGP'):
N, self.D = Y.shape
self.Y = Y
self.YYT = tdot(self.Y)
self.X = X
if kernels == None:
self.kern = []
for i in range(K):
self.kern.append(GPy.kern.RBF(input_dim=1))
else:
self.kern = kernels
CollapsedMixture.__init__(self, N, K, prior_Z, alpha, name)
self.link_parameter(
GPy.core.parameterization.param.Param(
'variance', variance,
GPy.core.parameterization.transformations.Logexp()))
self.link_parameters(*self.kern)
def __init__(self, X, Y, K=2, kernels=None, variance=1., alpha=1., prior_Z='symmetric', name='OMGP'):
N, self.D = Y.shape
self.Y = Y
self.X = X
if kernels == None:
self.kern = []
for i in range(K):
self.kern.append(GPy.kern.RBF(input_dim=1))
else:
self.kern = kernels
CollapsedMixture.__init__(self, N, K, prior_Z, alpha, name)
self.link_parameter(GPy.core.parameterization.param.Param('variance', variance, GPy.core.parameterization.transformations.Logexp()))
self.link_parameters(*self.kern)
def __init__(self, X, kernF, kernY, Y, K=2, alpha=1., prior_Z='symmetric', name='MOHGP'):
N,self.D = Y.shape
self.Y = Y
self.X = X
assert X.shape[0]==self.D, "input data don't match observations"
CollapsedMixture.__init__(self, N, K, prior_Z, alpha, name)
self.kernF = kernF
self.kernY = kernY
self.link_parameters(self.kernF, self.kernY)
#initialize kernels
self.Sf = self.kernF.K(self.X)
self.Sy = self.kernY.K(self.X)
self.Sy_inv, self.Sy_chol, self.Sy_chol_inv, self.Sy_logdet = pdinv(self.Sy+np.eye(self.D)*1e-6)
#Computations that can be done outside the optimisation loop
self.YYT = self.Y[:,:,np.newaxis]*self.Y[:,np.newaxis,:]
self.YTY = np.dot(self.Y.T,self.Y)
self.do_computations()
