import tensorflow as tf
import inferpy as inf
from inferpy.data import mnist
import tensorflow_probability as tfp
import numpy as np
N = 1000 # data size
(x_train, y_train), (x_test, y_test) = mnist.load_data(num_instances=N,
digits=[0, 1],
vectorize=False)
S = np.shape(x_train)[1:]
@inf.probmodel
def cnn_flipout_classifier(S):
with inf.datamodel():
x = inf.Normal(tf.ones(S), 1, name="x")
nn = inf.layers.Sequential([
tfp.layers.Convolution2DFlipout(4,
kernel_size=(10, 10),
padding="same",
activation="relu"),
tf.keras.layers.GlobalMaxPool2D(),
tf.keras.layers.Dense(1, activation='sigmoid')
])
y = inf.Normal(nn(tf.expand_dims(x, 1)), 0.001, name="y")
beta0 = inf.Normal(tf.ones([k, d0]) * loc_init, scale_init, name="beta0")
alpha0 = inf.Normal(tf.ones([d0]) * loc_init, scale_init, name="alpha0")
h0 = tf.nn.relu(z @ beta0 + alpha0, name="h0")
######
beta1 = inf.Normal(tf.ones([d0, 2*dx]) * loc_init, scale_init, name="beta1")
alpha1 = inf.Normal(tf.ones([2*dx]) * loc_init, scale_init, name="alpha1")
output = z @ beta0 + alpha0
return output
(x_train, y_train), _ = mnist.load_data(num_instances=N, digits=DIG)
m = nlpca(k, d0, dx, decoder)
inf_method = inf.inference.MCMC()
# learn the parameters
m.fit({"x": x_train}, inf_method)
# Plot the evolution of the loss
L = inf_method.losses
plt.plot(range(len(L)), L)
plt.xlabel('epochs')
plt.ylabel('Loss')
import tensorflow as tf
import inferpy as inf
from inferpy.data import mnist
import tensorflow_probability as tfp
N, M = 1000, 100 # data and batch size
(x_train, _), _ = mnist.load_data(num_instances=N, digits=[0, 1, 2])
# P model and the decoder NN
@inf.probmodel
def vae(k, d0, d, decoder):
with inf.datamodel():
z = inf.Normal(tf.ones(k), 1, name="z")
x = inf.Normal(decoder(d0, d, z), 1, name="x")
def decoder(d0, d, z):
return inf.layers.Sequential([
tfp.layers.DenseFlipout(d0, activation=tf.nn.relu),
tf.keras.layers.Dense(d),
])(z)
p = vae(k=2, d0=100, d=28 * 28, decoder=decoder)
# Q model and the encoder NN
@inf.probmodel
def qmodel(k, d0, d, encoder):
with inf.datamodel():