Bug fixes

Author: mnielsen

code/backprop2.py

@@ -42,9 +42,9 @@ def feedforward(self, a, start=0, end=None):
         """Return the result from feeding forward the activation ``a``
         from layer ``start`` through to layer ``end``.  Note that if
         ``end`` is ``None`` then this is interpreted as ``end =
-        self.num_layers``, i.e., the default behaviour is to propagate
+        self.num_layers-1``, i.e., the default behaviour is to propagate
         through to the end of the network."""
-        if end == None: end = self.num_layers
+        if end == None: end = self.num_layers-1
         for b, w in zip(self.biases, self.weights)[start:end]:
             a = sigmoid_vec(np.dot(w, a)+b)
         return a

code/deep_learning.py

@@ -30,22 +30,21 @@ def add_classifier_layer(net, num_outputs):
     return net_classifier
 
 def SGD_final_layer(
-    self, training_data, epochs, mini_batch_size, eta, lmbda, 
-    test=False, test_inputs=None, actual_test_results=None):
+    self, training_data, epochs, mini_batch_size, eta, lmbda):
     """
     Run SGD on the final layer of the Network ``self``.  Note that
     ``training_data`` is the input to the whole Network, not the
     encoded training data input to the final layer. 
     """
     encoded_training_data = [
-        (self.feedforward(x, start=0, end=-1), y) for x, y in training_data]
-    net = Network([self.sizes[-2], self.sizes[-1]])
-    net.biases[-1] = self.biases[-1]
-    net.weights[-1] = self.weights[-1]
-    net.SGD(encoded_training_data, epochs, mini_batch_size, eta,
-            lmbda, test, test_inputs, actual_test_results)
-    self.biases[-1] = net.biases[-1]
-    self.weights[-1] = net.weights[-1]
+        (self.feedforward(x, start=0, end=self.num_layers-2), y) 
+        for x, y in training_data]
+    net = Network(self.sizes[-2:])
+    net.biases[0] = self.biases[-1]
+    net.weights[0] = self.weights[-1]
+    net.SGD(encoded_training_data, epochs, mini_batch_size, eta, lmbda)
+    self.biases[-1] = net.biases[0]
+    self.weights[-1] = net.weights[0]
 
 
 # Add the SGD_final_layer method to the Network class