import numpy as np
from .. import inf
[docs]class model:
"""
Baysean Linear Model
Attributes
==========
prior: physbo.blm.prior.gauss
prior distribution of weights
lik: physbo.blm.lik.gauss
kernel
nbasis: int
number of features in random feature map
stats: Tuple
auxially parameters for sampling
method: str
sampling method
"""
def __init__(self, lik, prior, options={}):
self.prior = prior
self.lik = lik
self.nbasis = self.lik.linear.basis.nbasis
self._init_prior(prior)
self._set_options(options)
self.stats = ()
[docs] def prepare(self, X, t, Psi=None):
"""
initializes model by using the first training dataset
Parameters
==========
X: numpy.ndarray
inputs
t: numpy.ndarray
target (label)
Psi: numpy.ndarray
feature maps
See also
========
physbo.blm.inf.exact.prepare
"""
if self.method == "exact":
inf.exact.prepare(blm=self, X=X, t=t, Psi=Psi)
else:
pass
[docs] def update_stats(self, x, t, psi=None):
"""
updates model by using another training data
Parameters
==========
x: numpy.ndarray
input
t: float
target (label)
psi: numpy.ndarray
feature map
See also
========
physbo.blm.inf.exact.update_stats
"""
if self.method == "exact":
self.stats = inf.exact.update_stats(self, x, t, psi)
else:
pass
[docs] def get_post_params_mean(self):
"""
calculates posterior mean of weights
Returns
=======
numpy.ndarray
See also
========
physbo.blm.inf.exact.get_post_params_mean
"""
if self.method == "exact":
self.lik.linear.params = inf.exact.get_post_params_mean(blm=self)
[docs] def get_post_fmean(self, X, Psi=None, w=None):
"""
calculates posterior mean of model (function)
Parameters
==========
X: numpy.ndarray
inputs
Psi: numpy.ndarray
feature maps
w: numpy.ndarray
weight
See also
========
physbo.blm.inf.exact.get_post_fmean
"""
if self.method == "exact":
fmu = inf.exact.get_post_fmean(self, X, Psi, w)
else:
pass
return fmu
[docs] def sampling(self, w_mu=None, N=1, alpha=1.0):
"""
draws samples of weights
Parameters
==========
blm: physbo.blm.core.model
model
w_mu: numpy.ndarray
mean of weight
N: int
the number of samples
(default: 1)
alpha: float
noise for sampling source
(default: 1.0)
Returns
=======
numpy.ndarray
samples of weights
See also
========
physbo.blm.inf.exact.sampling
"""
if self.method == "exact":
w_hat = inf.exact.sampling(self, w_mu, N, alpha=alpha)
else:
pass
return w_hat
[docs] def post_sampling(self, Xtest, Psi=None, N=1, alpha=1.0):
"""
draws samples of mean value of model
Parameters
==========
Xtest: numpy.ndarray
inputs
Psi: numpy.ndarray
feature maps
(default: ``blm.lik.get_basis(Xtest)``)
N: int
number of samples
(default: 1)
alpha: float
noise for sampling source
Returns
=======
numpy.ndarray
"""
if Psi is None:
Psi = self.lik.get_basis(Xtest)
w_hat = self.sampling(N=N, alpha=alpha)
return Psi.dot(w_hat) + self.lik.linear.bias
[docs] def predict_sampling(self, Xtest, Psi=None, N=1):
"""
draws samples from model
Parameters
==========
Xtest: numpy.ndarray
inputs
Psi: numpy.ndarray
feature map
(default: ``blm.lik.get_basis(Xtest)``)
N: int
number of samples
(default: 1)
Returns
=======
numpy.ndarray
"""
if Xtest.shape[0] == 0:
return np.zeros((0, N))
fmean = self.post_sampling(Xtest, Psi, N=N)
A = np.random.randn(Xtest.shape[0], N)
return fmean + np.sqrt(self.lik.cov.sigma2) * A
[docs] def get_post_fcov(self, X, Psi=None, diag=True):
"""
calculates posterior covariance of model
Parameters
==========
X: numpy.ndarray
inputs
Psi: numpy.ndarray
feature maps
(default: blm.lik.linear.basis.get_basis(X))
diag: bool
if True, returns only variances as a diagonal matrix
(default: True)
Returns
=======
numpy.ndarray
See also
========
physbo.blm.inf.exact.get_post_fcov
"""
if self.method == "exact":
fcov = inf.exact.get_post_fcov(self, X, Psi, diag=True)
else:
pass
return fcov
def _set_options(self, options):
"""
read options
Parameters
==========
options: dict
- 'method' : sampling method
- 'exact' (default)
"""
self.method = options.get("method", "exact")
def _init_prior(self, prior):
"""
sets the prior distribution
Parameters
==========
prior: physbo.blm.prior.gauss
if None, prior.gauss(self.nbasis)
"""
if prior is None:
prior = prior.gauss(self.nbasis)
self.prior = prior