def resample(self, data=[], stats=None):
if len(data) > 0 or stats is not None:
stats = self._get_statistics(data) if stats is None else stats
for itr in range(self.niter):
self.A, self.sigma = \
sample_mniw(*self._natural_to_standard(
self.natural_hypparam + stats))
mat = self.M_0 - self.A
self.resample_K(1. / 2 * np.einsum(
'ij,ij->j', mat, np.linalg.solve(self.sigma, mat)))
else:
self.resample_K()
super(ARDRegression, self).resample()
def resample(self,data=[],stats=None):
if len(data) > 0 or stats is not None:
stats = self._get_statistics(data) if stats is None else stats
for itr in range(self.niter):
self.A, self.sigma = \
sample_mniw(*self._natural_to_standard(
self.natural_hypparam + stats))
mat = self.M_0 - self.A
self.resample_K(1./2*np.einsum(
'ij,ij->j',mat,np.linalg.solve(self.sigma,mat)))
else:
self.resample_K()
super(ARDRegression,self).resample()
def resample_from_mf(self):
self.A, self.sigma = sample_mniw(
*self._natural_to_standard(self.mf_natural_hypparam))
def resample(self, data=[], stats=None):
stats = self._get_statistics(data) if stats is None else stats
self.A, self.sigma = sample_mniw(
*self._natural_to_standard(self.natural_hypparam + stats))
self._initialize_mean_field()
def resample_from_mf(self):
self.A, self.sigma = sample_mniw(
*self._natural_to_standard(self.mf_natural_hypparam))
def resample(self,data=[],stats=None):
stats = self._get_statistics(data) if stats is None else stats
self.A, self.sigma = sample_mniw(
*self._natural_to_standard(self.natural_hypparam + stats))
self._initialize_mean_field()
