def sample_atom(self, data, cell):
a = self.base_measure.params.a
for item in cell.items:
a += data[item].x
n = cell.size
b = self.base_measure.params.b
b = b / (n * b + 1)
return GammaData(gamma_rvs(a, b))
def draw_from_prior(base_measure, size):
alpha = gamma_rvs(1, 1)
partition = sample_from_crp(alpha, size, base_measure)
# partition = Partition()
#
# partition.add_cell(base_measure.random())
#
# for item in range(size):
# partition.add_item(item, 0)
return alpha, partition
def sample(self, old_value, num_clusters, num_data_points):
a = self.a
b = self.b
k = num_clusters
n = num_data_points
eta = beta_rvs(old_value + 1, n)
x = (a + k - 1) / (n * (b - log(eta)))
pi = x / (1 + x)
label = discrete_rvs([pi, 1 - pi])
scale = b - log(eta)
if label == 0:
new_value = gamma_rvs(a + k, scale)
else:
new_value = gamma_rvs(a + k - 1, scale)
return new_value
def _sample_precision(self, sample_size, sample_mean, sample_variance):
prior_alpha = self.base_measure.params.alpha
prior_beta = self.base_measure.params.beta
prior_mean = self.base_measure.params.mean
prior_size = self.base_measure.params.size
posterior_alpha = prior_alpha + sample_size / 2
posterior_beta = prior_beta + \
(sample_size * sample_variance) / 2 + \
((prior_size * sample_size) / (2 * (prior_size + sample_size))) * (sample_mean - prior_mean) ** 2
return gamma_rvs(posterior_alpha, posterior_beta)
def random(self):
precision = gamma_rvs(self.params.alpha, self.params.beta) + 1e-10
mean = gaussian_rvs(self.params.mean, self.params.size * precision)
return GaussianGammaData(mean, precision)
def random(self):
x = gamma_rvs(self.params.a, self.params.b)
return GammaData(x)
def random(self, params):
a, b = self._get_params(params.x)
return GammaData(gamma_rvs(a, b))
