Refactoring

Author: mnielsen

code/autoencoder.py code/deep_autoencoder.pycode/autoencoder.py renamed to code/deep_autoencoder.py

@@ -1,6 +1,6 @@
 """
-autoencoder
-~~~~~~~~~~~
+deep_autoencoder
+~~~~~~~~~~~~~~~~
 
 A module which implements deep autoencoders.  
 """

code/mnist_autoencoder.py

@@ -2,8 +2,11 @@
 mnist_autoencoder
 ~~~~~~~~~~~~~~~~~
 
-Implements an autoencoder for the MNIST data, and plots the
-autoencoder's output for the first ten digits in the MNIST test set.
+Implements an autoencoder for the MNIST data.  The program can do two
+things: (1) plot the autoencoder's output for the first ten images in
+the MNIST test set; and (2) use the autoencoder to build a classifier.
+The program is a quick-and-dirty hack --- we'll do things in a more
+systematic way in the module ``deep_autoencoder``.
 """
 
 # My Libraries
@@ -15,14 +18,28 @@
 import matplotlib.pyplot as plt
 import numpy as np
 
-def mnist_autoencoder(hidden_units):
-    # Do the training
+def autoencoder_results(hidden_units):
+    """
+    Train an autoencoder using the MNIST training data and plot the
+    results when the first ten MNIST test images are passed through
+    the autoencoder.
+    """
     training_data, test_inputs, actual_test_results = \
         mnist_loader.load_data_nn()
+    net = train_autoencoder(hidden_units, training_data)
+    plot_test_results(net, test_inputs, actual_test_results)
+
+def train_autoencoder(hidden_units, training_data):
+    "Return a trained autoencoder."
     autoencoder_training_data = [(x, x) for x, _ in training_data]
     net = Network([784, hidden_units, 784])
     net.SGD(autoencoder_training_data, 3, 10, 0.01, 0.05)
-    # Plot the first ten test outputs
+    return net
+
+def plot_test_results(net, test_inputs, actual_test_results):
+    """
+    Plot the results after passing the first ten test MNIST digits through
+    the autoencoder ``net``."""
     fig = plt.figure()
     ax = fig.add_subplot(111)
     images_in = [test_inputs[j].reshape(-1, 28) for j in range(10)]
@@ -35,3 +52,19 @@ def mnist_autoencoder(hidden_units):
     plt.xticks(np.array([]))
     plt.yticks(np.array([]))
     plt.show()
+
+def classifier(hidden_units):
+    """
+    Train an autoencoder using the MNIST training data, and then use
+    the autoencoder to create a classifier with a single hidden layer.
+    """
+    training_data, test_inputs, actual_test_results = \
+        mnist_loader.load_data_nn()
+    net_ae = train_autoencoder(hidden_units, training_data)
+    net_c = Network([784, hidden_units, 10])
+    net_c.biases = net_ae.biases[:2]+[np.random.randn(10, 1)/np.sqrt(10)]
+    net_c.weights = net_ae.weights[:2]+\
+        [np.random.randn(10, hidden_units)/np.sqrt(10)]
+    net_c.SGD(training_data, 3, 10, 0.01, 0.05)
+    print net_c.evaluate(test_inputs, actual_test_results)
+    return net_c

code/mnist_pca.py

@@ -20,7 +20,7 @@
 pca = PCA(n_components=30)
 nn_images = [x for (x, y) in training_data]
 pca_images = np.concatenate(nn_images, axis=1).transpose()
-pca_r = pca.fit(pca_images[:1000])
+pca_r = pca.fit(pca_images)
 
 # Try PCA on first ten test images
 test_images = np.array(test_inputs[:10]).reshape((10,784))