# -*- coding: utf-8 -*-
import numpy as np
import tensorflow as tf
from distribution_base import DistributionBase
from full_covariance import FullCovariance
class GaussianDistribution(DistributionBase):
def __init__(self, dims, mean=None, covariance=None):
self.dims = dims
self.mean = mean
self.covariance = covariance
self.tf_mean = None
self.tf_covariance = None
self.tf_ln2piD = None
def initialize(self, dtype=tf.float64):
if self.tf_mean is None:
if self.mean is not None:
self.tf_mean = tf.Variable(self.mean, dtype=dtype)
else:
self.tf_mean = tf.Variable(tf.cast(tf.fill([self.dims], 0.0), dtype))
if self.tf_covariance is None:
if self.covariance is not None:
self.tf_covariance = self.covariance
else:
self.tf_covariance = FullCovariance(self.dims)
self.tf_covariance.initialize(dtype)
if self.tf_ln2piD is None:
self.tf_ln2piD = tf.constant(np.log(2 * np.pi) * self.dims, dtype=dtype)
def get_parameters(self):
return [
self.tf_mean,
self.tf_covariance.get_matrix()
]
def get_log_probabilities(self, data):
tf_quadratic_form = self.tf_covariance.get_inv_quadratic_form(data[0], self.tf_mean)
tf_log_coefficient = self.tf_ln2piD + self.tf_covariance.get_log_determinant()
return -0.5 * (tf_log_coefficient + tf_quadratic_form)
def get_parameter_updaters(self, data, gamma_weighted, gamma_sum):
tf_new_mean = tf.reduce_sum(data[0] * tf.expand_dims(gamma_weighted, 1), 0)
tf_covariance_updater = self.tf_covariance.get_value_updater(
data[0], tf_new_mean, gamma_weighted, gamma_sum)
return [tf_covariance_updater, self.tf_mean.assign(tf_new_mean)]
