import numpy as np
[docs]class gauss:
"""
Gaussian
Attributes
==========
linear
cov: blm.lik.cov
covariance
stats
"""
def __init__(self, linear, cov):
self.linear = linear
self.cov = cov
self.stats = ()
[docs] def get_cov(self, N, params=None):
"""
Returns covariance matrix
Parameters
==========
N: int
dimension
params: float
half of log of variance
(default: self.cov.params)
Returns
=======
numpy.ndarray
NxN covariance matrix
"""
if params is None:
params = np.copy(self.cov.params)
return self.cov.get_cov(N, params)
[docs] def get_prec(self, N, params=None):
"""
Returns precision matrix
Parameters
==========
N: int
dimension
params: float
half of log of variance
(default: self.cov.params)
Returns
=======
numpy.ndarray
NxN precision matrix
"""
if params is None:
params = np.copy(self.cov.params)
return self.cov.get_cov(N, params)
[docs] def get_basis(self, X):
"""
calculates value of basis function at input
Parameters
==========
X: numpy.ndarray
input
See also
========
blm.basis.fourier.get_basis
"""
return self.linear.basis.get_basis(X)
[docs] def get_mean(self, X, Psi=None, params=None, bias=None):
"""
calculates mean value
Parameters
==========
X: numpy.ndarray
raw input
Psi: numpy.ndarray
value of feature maps
params: numpy.ndarray
weight
bias: float
bias
See also
========
blm.basis.fourier.get_mean
"""
return self.linear.get_mean(X, Psi, params, bias)
[docs] def set_params(self, params):
"""
sets parameters
"""
self.linear.set_params(params)
[docs] def set_bias(self, bias):
"""
sets bias
"""
self.linear.set_bias(bias)
[docs] def sampling(self, fmean):
"""
draws samples
Parameters
==========
fmean: numpy.ndarray
means of samples
Returns
=======
samples: numpy.ndarray
"""
num_data = fmean.shape[0]
eps = np.sqrt(self.cov.sigma2) * np.random.randn(num_data)
return fmean + eps