first yadll commit

tests/test_layers.py

@@ -236,7 +236,7 @@ def input_data(self):
     @pytest.fixture
     def input_layer(self, input_data):
         from yadll.layers import InputLayer
-        shape = (10, 20)
+        shape = (None, 20)
         return InputLayer(shape, input_var=input_data)
 
     @pytest.fixture
@@ -310,6 +310,11 @@ def input_layer(self, input_data):
     def layer(self, pool_layer, input_layer):
         return pool_layer(input_layer, poolsize=(2, 2))
 
+    def test_output_shape(self, layer):
+        assert layer.output_shape() == (layer.input_shape[0],
+                                        layer.input_shape[1],
+                                        layer.input_shape[2] / layer.poolsize[0],
+                                        layer.input_shape[3] / layer.poolsize[1])
 
 class TestConvLayer:
     @pytest.fixture
@@ -339,6 +344,20 @@ def rbm(self):
         return RBM
 
 
+class TestBatchNormalization:
+    @pytest.fixture
+    def batch_normalization(self):
+        from yadll.layers import BatchNormalization
+        return BatchNormalization
+
+
+class TestLayerNormalization:
+    @pytest.fixture
+    def layer_normalization(self):
+        from yadll.layers import BatchNormalization
+        return BatchNormalization
+
+
 class TestRNN:
     @pytest.fixture
     def rnn(self):

yadll/layers.py

@@ -256,6 +256,10 @@ def __init__(self, incoming, corruption_level=0.5, **kwargs):
 
     def get_output(self, stochastic=True, **kwargs):
         X = self.input_layer.get_output(stochastic=stochastic, **kwargs)
+        if self.input_shape[0] is None:
+            lst = list(self.input_shape)
+            lst[0] = T.shape(X)[0]
+            self.input_shape = tuple(lst)
         if self.p != 1 and stochastic:
             X = X * T_rng.binomial(self.input_shape, n=1, p=self.p, dtype=floatX)
         return X
@@ -521,6 +525,43 @@ def get_unsupervised_cost(self, persistent=None, k=1, **kwargs):
         return monitoring_cost, updates
 
 
+class BatchNormalization(Layer):
+    """
+    Normalize the previous layer at each batch
+
+    References
+    ----------
+
+    ..[1] http://jmlr.org/proceedings/papers/v37/ioffe15.pdf
+    ..[2] https://github.com/fchollet/keras/blob/master/keras/layers/normalization.py#L6
+    """
+    def __init__(self, incoming, **kwargs):
+        super(BatchNormalization, self).__init__(incoming, **kwargs)
+
+    def get_output(self, **kwargs):
+        X = self.input_layer.get_output(**kwargs)
+        # TODO
+        return X
+
+
+class LayerNormalization(Layer):
+    """
+    Normalize the previous layer at each batch
+
+    References
+    ----------
+
+    ..[1] http://arxiv.org/pdf/1607.06450v1.pdf
+    """
+    def __init__(self, incoming, **kwargs):
+        super(LayerNormalization, self).__init__(incoming, **kwargs)
+
+    def get_output(self, **kwargs):
+        X = self.input_layer.get_output(**kwargs)
+        # TODO
+        return X
+
+
 class RNN(Layer):
     """
     Recurrent Neural Network