test_autoencoder.py

from __future__ import print_function
import numpy as np
from netlearner.autoencoder import MaskNoiseAutoencoder
from tensorflow.examples.tutorials.mnist import input_data
import matplotlib.pyplot as plt


mnist = input_data.read_data_sets('MNIST_data', one_hot=True)
train_dataset = mnist.train.images
test_dataset = mnist.test.images
print('Trainset', train_dataset.shape)
print('Testset', test_dataset.shape)
feature_size = train_dataset.shape[1]
encoder_size = 1000
autoencoder = MaskNoiseAutoencoder(
    feature_size, encoder_size, mask_frac=0.8, encode_lr=0.002)
batch_size = 100
num_steps = 50000
autoencoder.train(train_dataset, batch_size, num_steps)
test_loss = autoencoder.calc_reconstruct_loss(test_dataset)
print("Testset decode loss: %f" % test_loss)

rand_train = [train_dataset[8000]]
reconstruct_train = autoencoder.reconstruct(rand_train)
rand_train = np.multiply(255, np.reshape(rand_train, [28, 28]))
reconstruct_train = np.multiply(
    255, np.reshape(reconstruct_train, [28, 28]))

rand_test = [test_dataset[6000]]
reconstruct_test = autoencoder.reconstruct(rand_test)
rand_test = np.multiply(255, np.reshape(rand_test, [28, 28]))
reconstruct_test = np.multiply(255, np.reshape(reconstruct_test, [28, 28]))

plt.matshow(rand_test, cmap=plt.cm.gray)
plt.title('Random test image')
plt.figure(1)
plt.matshow(reconstruct_test, cmap=plt.cm.gray)
plt.title('Reconstructed test image')

plt.figure(2)
plt.matshow(rand_train, cmap=plt.cm.gray)
plt.title('Random train image')
plt.figure(3)
plt.matshow(reconstruct_train, cmap=plt.cm.gray)
plt.title('Reconstructed train image')
plt.show()