def sample_indicator(self, like, null_class_proba):
"""
sample the indicator from the likelihood
Parameters
----------
like: array of shape (nbitem,self.k)
component-wise likelihood
null_class_proba: array of shape(n_samples),
prior probability to be under the null
Returns
-------
z: array of shape(nbitem): a draw of the membership variable
Notes
-----
Here z=-1 encodes for the null class
"""
n = like.shape[0]
conditional_like_1 = ((1 - null_class_proba) * like.T).T
conditional_like_0 = np.reshape(null_class_proba *
self.null_dens, (n, 1))
conditional_like = np.hstack((conditional_like_0, conditional_like_1))
z = BGMM.sample_indicator(self, conditional_like) - 1
z[z == self.k] = self.k + np.arange(np.sum(z == self.k))
return z
def sample_indicator(self, like, null_class_proba):
"""
sample the indicator from the likelihood
Parameters
----------
like: array of shape (nbitem,self.k)
component-wise likelihood
null_class_proba: array of shape(n_samples),
prior probability to be under the null
Returns
-------
z: array of shape(nbitem): a draw of the membership variable
Notes
-----
Here z=-1 encodes for the null class
"""
n = like.shape[0]
conditional_like_1 = ((1 - null_class_proba) * like.T).T
conditional_like_0 = np.reshape(null_class_proba * self.null_dens,
(n, 1))
conditional_like = np.hstack((conditional_like_0, conditional_like_1))
z = BGMM.sample_indicator(self, conditional_like) - 1
z[z == self.k] = self.k + np.arange(np.sum(z == self.k))
return z
def update(self, x, z):
""" Update function (draw a sample of the IMM parameters)
Parameters
----------
x array of shape (n_samples,self.dim)
the data used in the estimation process
z array of shape (n_samples), type = np.int
the corresponding classification
"""
# re-dimension the priors in order to match self.k
self.prior_means = np.repeat(self._prior_means, self.k, 0)
self.prior_dof = self._prior_dof * np.ones(self.k)
self.prior_shrinkage = self._prior_shrinkage * np.ones(self.k)
self._dets = self._dets_ * np.ones(self.k)
self._inv_prior_scale = np.repeat(self._inv_prior_scale_, self.k, 0)
# initialize some variables
self.means = np.zeros((self.k, self.dim))
self.precisions = np.zeros((self.k, self.dim, self.dim))
# proceed with the update
BGMM.update(self, x, z)
def update(self, x, z):
""" Update function (draw a sample of the IMM parameters)
Parameters
----------
x array of shape (n_samples,self.dim)
the data used in the estimation process
z array of shape (n_samples), type = np.int
the corresponding classification
"""
# re-dimension the priors in order to match self.k
self.prior_means = np.repeat(self._prior_means, self.k, 0)
self.prior_dof = self._prior_dof * np.ones(self.k)
self.prior_shrinkage = self._prior_shrinkage * np.ones(self.k)
self._dets = self._dets_ * np.ones(self.k)
self._inv_prior_scale = np.repeat(self._inv_prior_scale_, self.k, 0)
# initialize some variables
self.means = np.zeros((self.k, self.dim))
self.precisions = np.zeros((self.k, self.dim, self.dim))
# proceed with the update
BGMM.update(self, x, z)
def sample_indicator(self, like):
""" Sample the indicator from the likelihood
Parameters
----------
like: array of shape (nbitem,self.k)
component-wise likelihood
Returns
-------
z: array of shape(nbitem): a draw of the membership variable
Notes
-----
The behaviour is different from standard bgmm in that z can take
arbitrary values
"""
z = BGMM.sample_indicator(self, like)
z[z == self.k] = self.k + np.arange(np.sum(z == self.k))
return z
def sample_indicator(self, like):
""" Sample the indicator from the likelihood
Parameters
----------
like: array of shape (nbitem,self.k)
component-wise likelihood
Returns
-------
z: array of shape(nbitem): a draw of the membership variable
Notes
-----
The behaviour is different from standard bgmm in that z can take
arbitrary values
"""
z = BGMM.sample_indicator(self, like)
z[z == self.k] = self.k + np.arange(np.sum(z == self.k))
return z