In [3]:

    

import matplotlib.pyplot as plt
import numpy as np
import tensorflow as tf
import math


    

In [4]:

    

NSAMPLE = 2500

y_data = np.float32(np.random.uniform(-10.5, 10.5, (1, NSAMPLE))).T
r_data = np.float32(np.random.normal(size=(NSAMPLE,1))) # random noise
x_data = np.float32(np.sin(0.75*y_data)*7.0+y_data*0.5+r_data*1.0)

plt.figure(figsize=(8, 8))
plt.plot(x_data,y_data,'ro', alpha=0.3)
plt.show()


    

In [5]:

    

NHIDDEN = 24
STDEV = 0.5
KMIX = 24 # number of mixtures
NOUT = KMIX * 3 # pi, mu, stdev

x = tf.placeholder(dtype=tf.float32, shape=[None,1], name="x")
y = tf.placeholder(dtype=tf.float32, shape=[None,1], name="y")

Wh = tf.Variable(tf.random_normal([1,NHIDDEN], stddev=STDEV, dtype=tf.float32))
bh = tf.Variable(tf.random_normal([1,NHIDDEN], stddev=STDEV, dtype=tf.float32))

Wo = tf.Variable(tf.random_normal([NHIDDEN,NOUT], stddev=STDEV, dtype=tf.float32))
bo = tf.Variable(tf.random_normal([1,NOUT], stddev=STDEV, dtype=tf.float32))

hidden_layer = tf.nn.tanh(tf.matmul(x, Wh) + bh)
output = tf.matmul(hidden_layer,Wo) + bo

def get_mixture_coef(output):
    out_pi = tf.placeholder(dtype=tf.float32, shape=[None,KMIX], name="mixparam")
    out_sigma = tf.placeholder(dtype=tf.float32, shape=[None,KMIX], name="mixparam")
    out_mu = tf.placeholder(dtype=tf.float32, shape=[None,KMIX], name="mixparam")

    out_pi, out_sigma, out_mu = tf.split(1, 3, output)

    max_pi = tf.reduce_max(out_pi, 1, keep_dims=True)
    out_pi = tf.sub(out_pi, max_pi)

    out_pi = tf.exp(out_pi)

    normalize_pi = tf.inv(tf.reduce_sum(out_pi, 1, keep_dims=True))
    out_pi = tf.mul(normalize_pi, out_pi)

    out_sigma = tf.exp(out_sigma)

    return out_pi, out_sigma, out_mu

out_pi, out_sigma, out_mu = get_mixture_coef(output)


    

In [7]:

    

x_test = np.float32(np.arange(-15,15,0.1))
NTEST = x_test.size
x_test = x_test.reshape(NTEST,1) # needs to be a matrix, not a vector

def get_pi_idx(x, pdf):
  N = pdf.size
  accumulate = 0
  for i in range(0, N):
    accumulate += pdf[i]
    if (accumulate >= x):
      return i
  print 'error with sampling ensemble'
  return -1

def generate_ensemble(out_pi, out_mu, out_sigma, M = 10):
  NTEST = x_test.size
  result = np.random.rand(NTEST, M) # initially random [0, 1]
  rn = np.random.randn(NTEST, M) # normal random matrix (0.0, 1.0)
  mu = 0
  std = 0
  idx = 0

  # transforms result into random ensembles
  for j in range(0, M):
    for i in range(0, NTEST):
      idx = get_pi_idx(result[i, j], out_pi[i])
      mu = out_mu[i, idx]
      std = out_sigma[i, idx]
      result[i, j] = mu + rn[i, j]*std
  return result


    

In [ ]:

    

out_pi_test, out_sigma_test, out_mu_test = sess.run(get_mixture_coef(output), feed_dict={x: x_test})

y_test = generate_ensemble(out_pi_test, out_mu_test, out_sigma_test)

plt.figure(figsize=(8, 8))
plt.plot(x_data,y_data,'ro', x_test,y_test,'bo',alpha=0.3)
plt.show()