def model_build(self):
dim_r = 4
dim_h_hidden = 128
dim_g_hidden = 128
x_context = tf.placeholder(tf.float32, shape=(None, self.dim))
y_context = tf.placeholder(tf.float32, shape=(None, 1))
x_target = tf.placeholder(tf.float32, shape=(None, self.dim))
y_target = tf.placeholder(tf.float32, shape=(None, 1))
neural_process = NeuralProcess(x_context, y_context, x_target,
y_target, dim_r, dim_h_hidden,
dim_g_hidden)
return neural_process
x_init = x[index_init]
y_init = y[index_init]
index_lef = np.setdiff1d(index_ori, index_init)
'''
dim_r = 4
#dim_z = 2
dim_h_hidden = 128
dim_g_hidden = 128
sess = tf.Session()
x_context = tf.placeholder(tf.float32, shape=(None, 2))
y_context = tf.placeholder(tf.float32, shape=(None, 1))
x_target = tf.placeholder(tf.float32, shape=(None, 2))
y_target = tf.placeholder(tf.float32, shape=(None, 1))
neural_process = NeuralProcess(x_context, y_context, x_target, y_target, dim_r,
dim_h_hidden, dim_g_hidden)
train_op_and_loss = neural_process.init_NP(learning_rate=0.001)
init = tf.global_variables_initializer()
sess.run(init)
n_iter = 10001
plot_freq = 1000
x_1_t = np.reshape(x_1, (-1, 1))
x_2_t = np.reshape(x_2, (-1, 1))
x_star = np.concatenate((x_1_t, x_2_t), axis=1)
# eps_value = np.random.normal(size=(n_draws, dim_r))
# epsilon = tf.constant(eps_value, dtype=tf.float32)
predict_op = neural_process.posterior_predict(x_init, y_init, x_star)
dim_r = 2
#dim_z = 2
dim_h_hidden = 8
dim_g_hidden = 8
sess = tf.Session()
x_context = tf.placeholder(tf.float32, shape=(None, 1))
y_context = tf.placeholder(tf.float32, shape=(None, 1))
x_target = tf.placeholder(tf.float32, shape=(None, 1))
y_target = tf.placeholder(tf.float32, shape=(None, 1))
#neural_process = NeuralProcess(x_context, y_context, x_target, y_target,
# dim_r, dim_z, dim_h_hidden, dim_g_hidden)
neural_process = NeuralProcess(x_context, y_context, x_target, y_target, dim_r,
dim_h_hidden, dim_g_hidden)
train_op_and_loss = neural_process.init_NP(learning_rate=0.001)
init = tf.global_variables_initializer()
sess.run(init)
n_iter = 5000
plot_freq = 200
n_draws = 50
x_star_temp = np.linspace(x_min, x_max, n_f * 10)
x_star = np.expand_dims(x_star_temp, axis=1)
#eps_value = np.random.normal(size=(n_draws, dim_r))
eps_value = np.random.normal(size=(n_draws, 1))
epsilon = tf.constant(eps_value, dtype=tf.float32)