bernoulli_beta.py

import numpy as np
import tensorflow as tf
import edward as ed
from edward.models import Bernoulli, Beta, PointMass

ed.set_seed(42)

# DATA

n = 4
x_train = np.array([0, 1, 1, 1]).reshape((n,1))

# MODEL

alpha = 1.0
beta = 1.0
p = Beta(concentration1=tf.Variable([alpha]),concentration0=tf.Variable([beta]))
x = Bernoulli(probs=p, sample_shape=(n))

# INFERENCE by MAP

qp = PointMass(params=tf.Variable([0.1]))
inference = ed.MAP({p: qp}, data={x: x_train})
inference.run()

# CRITICISM
sess = ed.get_session()
print(sess.run(qp.mean()))
print(sess.run(qp.params))
print(qp.eval())

x_post = ed.copy(x, {p: qp})
print(ed.evaluate('binary_accuracy', data={x_post: x_train}))

# INFERENCE by KLqp
alpha = 1.0
beta = 1.0
qp = Beta(concentration1=tf.Variable([alpha]),concentration0=tf.Variable([beta]))

inference = ed.KLqp({p: qp}, data={x: x_train})
inference.run()

# CRITICISM

sess = ed.get_session()
print(sess.run(qp.mean()))
print(sess.run(qp.concentration0), sess.run(qp.concentration1))
print(qp.eval())

x_post = ed.copy(x, {p: qp})
print(ed.evaluate('binary_accuracy', data={x_post: x_train}))