remove unnecessary floating point casts in examples

docs/tex/iclr2017.tex

@@ -111,7 +111,7 @@ \subsubsection{Section 4. Compositional Representations for Inference}
 import tensorflow as tf
 from edward.models import Categorical, Normal
 
-x_train = np.zeros([N, D], dtype=np.float32)
+x_train = np.zeros([N, D])
 
 qbeta = Normal(mu=tf.Variable(tf.zeros([K, D])),
                sigma=tf.exp(tf.Variable(tf.zeros([K, D]))))
@@ -128,7 +128,7 @@ \subsubsection{Section 4. Compositional Representations for Inference}
 import tensorflow as tf
 from edward.models import Empirical
 
-x_train = np.zeros([N, D], dtype=np.float32)
+x_train = np.zeros([N, D])
 
 T = 10000  # number of samples
 qbeta = Empirical(params=tf.Variable(tf.zeros([T, K, D])))
@@ -158,7 +158,7 @@ \subsubsection{Section 4. Compositional Representations for Inference}
   return Dense(1, activation=None)(h)
 
 # DATA
-x_train = np.zeros([N, 28 * 28], dtype=np.float32)
+x_train = np.zeros([N, 28 * 28])
 
 # MODEL
 eps = Normal(mu=tf.zeros([N, d]), sigma=tf.ones([N, d]))
@@ -181,7 +181,7 @@ \subsubsection{Section 4. Compositional Representations for Inference}
 from edward.models import Categorical, PointMass
 
 # DATA
-x_train = np.zeros([N, D], dtype=np.float32)
+x_train = np.zeros([N, D])
 
 # INFERENCE
 qbeta = PointMass(params=tf.Variable(tf.zeros([K, D])))
@@ -253,8 +253,8 @@ \subsubsection{Section 5. Experiments}
 T = 100  # number of empirical samples
 
 # DATA
-x_data = np.zeros([N, D], dtype=np.float32)
-y_data = np.zeros([N], dtype=np.float32)
+x_data = np.zeros([N, D])
+y_data = np.zeros([N])
 
 # MODEL
 x = tf.Variable(x_data, trainable=False)

docs/tex/tutorials/mixture-density-network.tex

@@ -21,8 +21,8 @@ \subsubsection{Data}
 from sklearn.model_selection import train_test_split
 
 def build_toy_dataset(N):
-  y_data = np.random.uniform(-10.5, 10.5, N).astype(np.float32)
-  r_data = np.random.normal(size=N).astype(np.float32)  # random noise
+  y_data = np.random.uniform(-10.5, 10.5, N)
+  r_data = np.random.normal(size=N)  # random noise
   x_data = np.sin(0.75 * y_data) * 7.0 + y_data * 0.5 + r_data * 1.0
   x_data = x_data.reshape((N, 1))
   return train_test_split(x_data, y_data, random_state=42)

docs/tex/tutorials/supervised-classification.tex

@@ -16,7 +16,7 @@ \subsubsection{Data}
 {https://stat.ethz.ch/R-manual/R-devel/library/MASS/html/crabs.html}
 {crabs data set}.
 \begin{lstlisting}[language=Python]
-df = np.loadtxt('data/crabs_train.txt', dtype='float32', delimiter=',')
+df = np.loadtxt('data/crabs_train.txt', delimiter=',')
 df[df[:, 0] == -1, 0] = 0  # replace -1 label with 0 label
 
 N = 25  # number of data points

docs/tex/tutorials/supervised-regression.tex

@@ -20,7 +20,7 @@ \subsubsection{Data}
 \begin{lstlisting}[language=Python]
 def build_toy_dataset(N, w, noise_std=0.1):
   D = len(w)
-  x = np.random.randn(N, D).astype(np.float32)
+  x = np.random.randn(N, D)
   y = np.dot(x, w) + np.random.normal(0, noise_std, size=N)
   return x, y
 
@@ -135,7 +135,7 @@ \subsubsection{Criticism}
   w_samples = w.sample(n_samples).eval()
   b_samples = b.sample(n_samples).eval()
   plt.scatter(X_data, y_data)
-  inputs = np.linspace(-1, 10, num=400, dtype=np.float32)
+  inputs = np.linspace(-1, 10, num=400)
   for ns in range(n_samples):
     output = inputs * w_samples[ns] + b_samples[ns]
     plt.plot(inputs, output)

docs/tex/tutorials/unsupervised.tex

@@ -18,7 +18,7 @@ \subsubsection{Data}
   pi = np.array([0.4, 0.6])
   mus = [[1, 1], [-1, -1]]
   stds = [[0.1, 0.1], [0.1, 0.1]]
-  x = np.zeros((N, 2), dtype=np.float32)
+  x = np.zeros((N, 2))
   for n in range(N):
     k = np.argmax(np.random.multinomial(1, pi))
     x[n, :] = np.random.multivariate_normal(mus[k], np.diag(stds[k]))

examples/bayesian_linear_regression_implicitklqp.py

@@ -30,7 +30,7 @@
 
 def build_toy_dataset(N, w, noise_std=0.1):
   D = len(w)
-  x = np.random.randn(N, D).astype(np.float32)
+  x = np.random.randn(N, D)
   y = np.dot(x, w) + np.random.normal(0, noise_std, size=N)
   return x, y
 

examples/bayesian_linear_regression_ppc.py

@@ -23,8 +23,7 @@ def build_toy_dataset(N, noise_std=0.5):
   X = np.concatenate([np.linspace(0, 2, num=N / 2),
                       np.linspace(6, 8, num=N / 2)])
   y = 2.0 * X + 10 * np.random.normal(0, noise_std, size=N)
-  X = X.astype(np.float32).reshape((N, 1))
-  y = y.astype(np.float32)
+  X = X.reshape((N, 1))
   return X, y
 
 
@@ -68,7 +67,7 @@ def build_toy_dataset(N, noise_std=0.5):
 
 plt.scatter(X_train, y_train)
 
-inputs = np.linspace(-1, 10, num=400, dtype=np.float32)
+inputs = np.linspace(-1, 10, num=400)
 for ns in range(n_prior_samples):
     output = inputs * w_prior[ns] + b_prior[ns]
     plt.plot(inputs, output)
@@ -83,7 +82,7 @@ def build_toy_dataset(N, noise_std=0.5):
 
 plt.scatter(X_train, y_train)
 
-inputs = np.linspace(-1, 10, num=400, dtype=np.float32)
+inputs = np.linspace(-1, 10, num=400)
 for ns in range(n_posterior_samples):
     output = inputs * w_post[ns] + b_post[ns]
     plt.plot(inputs, output)

examples/bayesian_linear_regression_sghmc.py

@@ -24,8 +24,7 @@ def build_toy_dataset(N, noise_std=0.5):
   X = np.concatenate([np.linspace(0, 2, num=N / 2),
                       np.linspace(6, 8, num=N / 2)])
   y = 2.0 * X + 10 * np.random.normal(0, noise_std, size=N)
-  X = X.astype(np.float32).reshape((N, 1))
-  y = y.astype(np.float32)
+  X = X.reshape((N, 1))
   return X, y
 
 
@@ -78,7 +77,7 @@ def build_toy_dataset(N, noise_std=0.5):
 
 plt.scatter(X_train, y_train)
 
-inputs = np.linspace(-1, 10, num=400, dtype=np.float32)
+inputs = np.linspace(-1, 10, num=400)
 for ns in range(n_prior_samples):
     output = inputs * w_prior[ns] + b_prior[ns]
     plt.plot(inputs, output)
@@ -93,7 +92,7 @@ def build_toy_dataset(N, noise_std=0.5):
 
 plt.scatter(X_train, y_train)
 
-inputs = np.linspace(-1, 10, num=400, dtype=np.float32)
+inputs = np.linspace(-1, 10, num=400)
 for ns in range(n_posterior_samples):
     output = inputs * w_post[ns] + b_post[ns]
     plt.plot(inputs, output)

examples/bayesian_linear_regression_tensor.py

@@ -24,8 +24,7 @@ def build_toy_dataset(N, noise_std=0.1):
   X = np.concatenate([np.linspace(0, 2, num=N / 2),
                       np.linspace(6, 8, num=N / 2)])
   y = 5.0 * X + np.random.normal(0, noise_std, size=N)
-  X = X.astype(np.float32).reshape((N, 1))
-  y = y.astype(np.float32)
+  X = X.reshape((N, 1))
   return X, y
 
 
@@ -38,7 +37,7 @@ def build_toy_dataset(N, noise_std=0.1):
 X_data, y_data = build_toy_dataset(N)
 
 # MODEL
-X = X_data
+X = tf.cast(X_data, tf.float32)
 w = Normal(mu=tf.zeros(D), sigma=tf.ones(D))
 b = Normal(mu=tf.zeros(1), sigma=tf.ones(1))
 y = Normal(mu=ed.dot(X, w) + b, sigma=tf.ones(N))

examples/bayesian_nn.py

@@ -22,17 +22,16 @@
 
 def build_toy_dataset(N=40, noise_std=0.1):
   D = 1
-  x = np.concatenate([np.linspace(0, 2, num=N / 2),
+  X = np.concatenate([np.linspace(0, 2, num=N / 2),
                       np.linspace(6, 8, num=N / 2)])
-  y = np.cos(x) + np.random.normal(0, noise_std, size=N)
-  x = (x - 4.0) / 4.0
-  x = x.astype(np.float32).reshape((N, D))
-  y = y.astype(np.float32)
-  return x, y
+  y = np.cos(X) + np.random.normal(0, noise_std, size=N)
+  X = (X - 4.0) / 4.0
+  X = X.reshape((N, D))
+  return X, y
 
 
-def neural_network(x):
-  h = tf.tanh(tf.matmul(x, W_0) + b_0)
+def neural_network(X):
+  h = tf.tanh(tf.matmul(X, W_0) + b_0)
   h = tf.tanh(tf.matmul(h, W_1) + b_1)
   h = tf.matmul(h, W_2) + b_2
   return tf.reshape(h, [-1])
@@ -44,7 +43,7 @@ def neural_network(x):
 D = 1   # number of features
 
 # DATA
-x_train, y_train = build_toy_dataset(N)
+X_train, y_train = build_toy_dataset(N)
 
 # MODEL
 W_0 = Normal(mu=tf.zeros([D, 10]), sigma=tf.ones([D, 10]))
@@ -54,8 +53,8 @@ def neural_network(x):
 b_1 = Normal(mu=tf.zeros(10), sigma=tf.ones(10))
 b_2 = Normal(mu=tf.zeros(1), sigma=tf.ones(1))
 
-x = x_train
-y = Normal(mu=neural_network(x), sigma=0.1 * tf.ones(N))
+X = tf.placeholder(tf.float32, [N, D])
+y = Normal(mu=neural_network(X), sigma=0.1 * tf.ones(N))
 
 # INFERENCE
 qW_0 = Normal(mu=tf.Variable(tf.random_normal([D, 10])),
@@ -73,5 +72,5 @@ def neural_network(x):
 
 inference = ed.KLqp({W_0: qW_0, b_0: qb_0,
                      W_1: qW_1, b_1: qb_1,
-                     W_2: qW_2, b_2: qb_2}, data={y: y_train})
+                     W_2: qW_2, b_2: qb_2}, data={X: X_train, y: y_train})
 inference.run()

examples/dirichlet_categorical.py

@@ -16,26 +16,19 @@
 N = 1000
 K = 4
 
-# Data generation
-alpha = np.array([20., 30., 10., 10.])
-pi = np.random.dirichlet(alpha).astype(np.float32)
-zn_data = np.array([np.random.choice(K, 1, p=pi)[0] for n in range(N)])
-print('pi={}'.format(pi))
-
-# Prior definition
-alpha_prior = tf.Variable(np.array([1., 1., 1., 1.]),
-                          dtype=tf.float32, trainable=False)
-
-# Posterior inference
-# Probabilistic model
-pi = Dirichlet(alpha=alpha_prior)
-zn = Categorical(p=tf.ones([N, 1]) * pi)
-
-# Variational model
+# DATA
+pi_true = np.random.dirichlet(np.array([20.0, 30.0, 10.0, 10.0]))
+z_data = np.array([np.random.choice(K, 1, p=pi_true)[0] for n in range(N)])
+print('pi={}'.format(pi_true))
+
+# MODEL
+pi = Dirichlet(alpha=tf.ones(4))
+z = Categorical(p=tf.ones([N, 1]) * pi)
+
+# INFERENCE
 qpi = Dirichlet(alpha=tf.nn.softplus(tf.Variable(tf.random_normal([K]))))
 
-# Inference
-inference = ed.KLqp({pi: qpi}, data={zn: zn_data})
+inference = ed.KLqp({pi: qpi}, data={z: z_data})
 inference.run(n_iter=1500, n_samples=30)
 
 sess = ed.get_session()

examples/invgamma_normal_mh.py

@@ -22,8 +22,8 @@
 print('sigma={}'.format(sigma))
 
 # Prior definition
-alpha = tf.Variable(0.5, dtype=tf.float32, trainable=False)
-beta = tf.Variable(0.7, dtype=tf.float32, trainable=False)
+alpha = tf.Variable(0.5, trainable=False)
+beta = tf.Variable(0.7, trainable=False)
 
 # Posterior inference
 # Probabilistic model

examples/mixture_gaussian_mh.py

@@ -29,7 +29,7 @@ def build_toy_dataset(N):
   pi = np.array([0.4, 0.6])
   mus = [[1, 1], [-1, -1]]
   stds = [[0.1, 0.1], [0.1, 0.1]]
-  x = np.zeros((N, 2), dtype=np.float32)
+  x = np.zeros((N, 2))
   for n in range(N):
     k = np.argmax(np.random.multinomial(1, pi))
     x[n, :] = np.random.multivariate_normal(mus[k], np.diag(stds[k]))

examples/normal_normal.py

@@ -13,7 +13,7 @@
 ed.set_seed(42)
 
 # DATA
-x_data = np.array([0.0] * 50, dtype=np.float32)
+x_data = np.array([0.0] * 50)
 
 # MODEL: Normal-Normal with known variance
 mu = Normal(mu=0.0, sigma=1.0)

examples/normal_normal_hmc.py

@@ -14,7 +14,7 @@
 ed.set_seed(42)
 
 # DATA
-x_data = np.array([0.0] * 50, dtype=np.float32)
+x_data = np.array([0.0] * 50)
 
 # MODEL: Normal-Normal with known variance
 mu = Normal(mu=0.0, sigma=1.0)

examples/normal_normal_mh.py

@@ -14,7 +14,7 @@
 ed.set_seed(42)
 
 # DATA
-x_data = np.array([0.0] * 50, dtype=np.float32)
+x_data = np.array([0.0] * 50)
 
 # MODEL: Normal-Normal with known variance
 mu = Normal(mu=0.0, sigma=1.0)

examples/normal_normal_tensorboard.py

@@ -13,7 +13,7 @@
 ed.set_seed(42)
 
 # DATA
-x_data = np.array([0.0] * 50, dtype=np.float32)
+x_data = np.array([0.0] * 50)
 
 # MODEL: Normal-Normal with known variance
 mu = Normal(mu=0.0, sigma=1.0, name='mu')

examples/normal_sghmc.py

@@ -7,15 +7,14 @@
 from __future__ import print_function
 
 import edward as ed
-import tensorflow as tf
 import numpy as np
+import tensorflow as tf
+
 from matplotlib import pyplot as plt
 from edward.models import Empirical, MultivariateNormalFull
 
 plt.style.use("ggplot")
 
-# Plotting helper function.
-
 
 def mvn_plot_contours(z, label=False, ax=None):
   """
@@ -32,7 +31,7 @@ def mvn_plot_contours(z, label=False, ax=None):
   xs = np.linspace(xmin, xmax, num=100)
   ys = np.linspace(ymin, ymax, num=100)
   X, Y = np.meshgrid(xs, ys)
-  T = tf.convert_to_tensor(np.c_[X.flatten(), Y.flatten()], dtype=tf.float32)
+  T = tf.cast(np.c_[X.flatten(), Y.flatten()], dtype=tf.float32)
   Z = sess.run(tf.exp(z.log_prob(T))).reshape((len(xs), len(ys)))
   if ax is None:
     fig, ax = plt.subplots()
@@ -41,7 +40,6 @@ def mvn_plot_contours(z, label=False, ax=None):
     plt.clabel(cs, inline=1, fontsize=10)
 
 
-# Example body.
 ed.set_seed(42)
 
 # MODEL
@@ -62,7 +60,6 @@ def mvn_plot_contours(z, label=False, ax=None):
 print("Inferred posterior std:")
 print(std)
 
-# VISUALIZATION
 fig, ax = plt.subplots()
 trace = sess.run(qz.params)
 ax.scatter(trace[:, 0], trace[:, 1], marker=".")

examples/probabilistic_matrix_factorization.py

@@ -30,8 +30,8 @@ def get_indicators(N, M, prob_std=0.5):
 D = 3  # number of latent factors
 
 # true latent factors
-U_true = np.random.randn(D, N).astype(np.float32)
-V_true = np.random.randn(D, M).astype(np.float32)
+U_true = np.random.randn(D, N)
+V_true = np.random.randn(D, M)
 
 # DATA
 R_true = build_toy_dataset(U_true, V_true, N, M)