Layer API refactor

keras/layers/advanced_activations.py

@@ -6,7 +6,7 @@ class LeakyReLU(Layer):
         super(LeakyReLU,self).__init__()
         self.alpha = alpha
 
-    def output(self, train):
+    def get_output(self, train):
         X = self.get_input(train)
         return ((X + abs(X)) / 2.0) + self.alpha * ((X - abs(X)) / 2.0)
 
@@ -27,7 +27,7 @@ class PReLU(Layer):
         self.params = [self.alphas]
         self.input_shape = input_shape
 
-    def output(self, train):
+    def get_output(self, train):
         X = self.get_input(train)
         pos = ((X + abs(X)) / 2.0)
         neg = self.alphas * ((X - abs(X)) / 2.0)

keras/layers/convolutional.py

@@ -41,7 +41,7 @@ class Convolution2D(Layer):
         if weights is not None:
             self.set_weights(weights)
 
-    def output(self, train):
+    def get_output(self, train):
         X = self.get_input(train)
 
         conv_out = theano.tensor.nnet.conv.conv2d(X, self.W, 
@@ -69,7 +69,7 @@ class MaxPooling2D(Layer):
         self.poolsize = poolsize
         self.ignore_border = ignore_border
 
-    def output(self, train):
+    def get_output(self, train):
         X = self.get_input(train)
         output = downsample.max_pool_2d(X, self.poolsize, ignore_border=self.ignore_border)
         return output

keras/layers/core.py

@@ -16,15 +16,15 @@ class Layer(object):
     def __init__(self):
         self.params = []
 
-    def connect(self, previous_layer):
-        self.previous_layer = previous_layer
+    def connect(self, node):
+        self.previous = node
 
-    def output(self, train):
+    def get_output(self, train):
         raise NotImplementedError
 
     def get_input(self, train):
-        if hasattr(self, 'previous_layer'):
-            return self.previous_layer.output(train=train)
+        if hasattr(self, 'previous'):
+            return self.previous.get_output(train=train)
         else:
             return self.input
 
@@ -50,7 +50,7 @@ class Dropout(Layer):
         super(Dropout,self).__init__()
         self.p = p
 
-    def output(self, train):
+    def get_output(self, train):
         X = self.get_input(train)
         if self.p > 0.:
             retain_prob = 1. - self.p
@@ -75,7 +75,7 @@ class Activation(Layer):
         self.target = target
         self.beta = beta
 
-    def output(self, train):
+    def get_output(self, train):
         X = self.get_input(train)
         return self.activation(X)
 
@@ -96,7 +96,7 @@ class Reshape(Layer):
         super(Reshape,self).__init__()
         self.dims = dims
 
-    def output(self, train):
+    def get_output(self, train):
         X = self.get_input(train)
         nshape = make_tuple(X.shape[0], *self.dims)
         return theano.tensor.reshape(X, nshape)
@@ -114,7 +114,7 @@ class Flatten(Layer):
     def __init__(self):
         super(Flatten,self).__init__()
 
-    def output(self, train):
+    def get_output(self, train):
         X = self.get_input(train)
         size = theano.tensor.prod(X.shape) // X.shape[0]
         nshape = (X.shape[0], size)
@@ -132,7 +132,7 @@ class RepeatVector(Layer):
         super(RepeatVector,self).__init__()
         self.n = n
 
-    def output(self, train):
+    def get_output(self, train):
         X = self.get_input(train)
         tensors = [X]*self.n
         stacked = theano.tensor.stack(*tensors)
@@ -168,7 +168,7 @@ class Dense(Layer):
         if weights is not None:
             self.set_weights(weights)
 
-    def output(self, train):
+    def get_output(self, train):
         X = self.get_input(train)
         output = self.activation(T.dot(X, self.W) + self.b)
         return output
@@ -210,7 +210,7 @@ class TimeDistributedDense(Layer):
         if weights is not None:
             self.set_weights(weights)
 
-    def output(self, train):
+    def get_output(self, train):
         X = self.get_input(train)
 
         def act_func(X):

keras/layers/embeddings.py

@@ -27,7 +27,7 @@ class Embedding(Layer):
         if weights is not None:
             self.set_weights(weights)
 
-    def output(self, train=False):
+    def get_output(self, train=False):
         X = self.get_input(train)
         out = self.W[X]
         return out
@@ -83,7 +83,7 @@ class WordContextProduct(Layer):
             self.set_weights(weights)
 
 
-    def output(self, train=False):
+    def get_output(self, train=False):
         X = self.get_input(train)
         w = self.W_w[X[:, 0]] # nb_samples, proj_dim
         c = self.W_c[X[:, 1]] # nb_samples, proj_dim

keras/layers/normalization.py

@@ -27,7 +27,7 @@ class BatchNormalization(Layer):
         if weights is not None:
             self.set_weights(weights)
 
-    def output(self, train):
+    def get_output(self, train):
         X = self.get_input(train)
 
         if self.mode == 0:

keras/layers/recurrent.py

@@ -46,7 +46,7 @@ class SimpleRNN(Layer):
         '''
         return self.activation(x_t + T.dot(h_tm1, u))
 
-    def output(self, train):
+    def get_output(self, train):
         X = self.get_input(train) # shape: (nb_samples, time (padded with zeros at the end), input_dim)
         # new shape: (time, nb_samples, input_dim) -> because theano.scan iterates over main dimension
         X = X.dimshuffle((1,0,2)) 
@@ -119,7 +119,7 @@ class SimpleDeepRNN(Layer):
             o += self.inner_activation(T.dot(args[i], args[i+self.depth]))
         return self.activation(o)
 
-    def output(self, train):
+    def get_output(self, train):
         X = self.get_input(train)
         X = X.dimshuffle((1,0,2)) 
 
@@ -222,7 +222,7 @@ class GRU(Layer):
         h_t = z * h_tm1 + (1 - z) * hh_t
         return h_t
 
-    def output(self, train):
+    def get_output(self, train):
         X = self.get_input(train) 
         X = X.dimshuffle((1,0,2)) 
 
@@ -331,7 +331,7 @@ class LSTM(Layer):
         h_t = o_t * self.activation(c_t)
         return h_t, c_t
 
-    def output(self, train):
+    def get_output(self, train):
         X = self.get_input(train) 
         X = X.dimshuffle((1,0,2))
 

keras/models.py

@@ -66,7 +66,9 @@ class Sequential(object):
             self.constraints += [layer.constraint for _ in range(len(layer.params))]
         else:
             self.constraints += [constraints.identity for _ in range(len(layer.params))]
-        
+
+    def get_output(self, train):
+        return self.layers[-1].get_output(train)
 
     def compile(self, optimizer, loss, class_mode="categorical", y_dim_components=1):
         self.optimizer = optimizers.get(optimizer)
@@ -81,8 +83,8 @@ class Sequential(object):
             self.layers[0].input = ndim_tensor(ndim)
         self.X = self.layers[0].input
 
-        self.y_train = self.layers[-1].output(train=True)
-        self.y_test = self.layers[-1].output(train=False)
+        self.y_train = self.get_output(train=True)
+        self.y_test = self.get_output(train=False)
 
         # output of model
         self.y = ndim_tensor(y_dim_components+1)
@@ -114,6 +116,7 @@ class Sequential(object):
         self._test_with_acc = theano.function([self.X, self.y], [test_score, test_accuracy], 
             allow_input_downcast=True)
 
+
     def train(self, X, y, accuracy=False):
         y = standardize_y(y)
         if accuracy: