Merge remote-tracking branch 'upstream/master' into LiveMonitorGUI

Author: AdamStone

.requirements.txt

@@ -0,0 +1,4 @@
+# This is for readthedocs.org to be able to build the documentation,
+# ignore otherwise
+numpydoc
+git+git://github.com/Theano/Theano.git

pylearn2/datasets/dense_design_matrix.py

@@ -753,7 +753,11 @@ def set_topological_view(self, V, axes=('b', 0, 1, 'c')):
         V : ndarray
             An array containing a design matrix representation of
             training examples.
-        axes : WRITEME
+        axes : tuple, optional
+            The axes ordering of the provided topo_view. Must be some
+            permutation of ('b', 0, 1, 'c') where 'b' indicates the axis
+            indexing examples, 0 and 1 indicate the row/cols dimensions and
+            'c' indicates the axis indexing color channels.
         """
         if len(V.shape) != len(axes):
             raise ValueError("The topological view must have exactly 4 "
@@ -1169,11 +1173,14 @@ def set_topological_view(self, V, axes=('b', 0, 1, 'c'), start=0):
         ----------
         V : ndarray
             An array containing a design matrix representation of training \
-            examples. If unspecified, the entire dataset (`self.X`) is used \
-            instead.
-        axes : WRITEME
-            WRITEME
-        start : WRITEME
+            examples.
+        axes : tuple, optional
+            The axes ordering of the provided topo_view. Must be some
+            permutation of ('b', 0, 1, 'c') where 'b' indicates the axis
+            indexing examples, 0 and 1 indicate the row/cols dimensions and
+            'c' indicates the axis indexing color channels.
+        start : int
+            The start index to write data.
         """
         assert not contains_nan(V)
         rows = V.shape[axes.index(0)]
@@ -1189,11 +1196,15 @@ def set_topological_view(self, V, axes=('b', 0, 1, 'c'), start=0):
 
     def init_hdf5(self, path, shapes):
         """
-        .. todo::
+        Initializes the hdf5 file into which the data will be stored. This must
+        be called before calling fill_hdf5.
 
-            WRITEME properly
-
-        Initialize hdf5 file to be used ba dataset
+        Parameters
+        ----------
+        path : string
+            The name of the hdf5 file.
+        shapes : tuple
+            The shapes of X and y.
         """
 
         x_shape, y_shape = shapes
@@ -1217,14 +1228,25 @@ def fill_hdf5(file_handle,
                   start=0,
                   batch_size=5000):
         """
-        .. todo::
+        Saves the data to the hdf5 file.
 
-            WRITEME properly
-
-        PyTables tends to crash if you write large data on them at once.
-        This function write data on file_handle in batches
+        PyTables tends to crash if you write large amounts of data into them
+        at once. As such this function writes data in batches.
 
-        start: the start index to write data
+        Parameters
+        ----------
+        file_handle : hdf5 file handle
+            Handle to an hdf5 object.
+        data_x : nd array
+            X data. Must be the same shape as specified to init_hdf5.
+        data_y : nd array, optional
+            y data. Must be the same shape as specified to init_hdf5.
+        node : string, optional
+            The hdf5 node into which the data should be stored.
+        start : int
+            The start index to write data.
+        batch_size : int, optional
+            The size of the batch to be saved.
         """
 
         if node is None:
@@ -1245,9 +1267,17 @@ def fill_hdf5(file_handle,
 
     def resize(self, h5file, start, stop):
         """
-        .. todo::
+        Resizes the X and y tables. This must be called before calling
+        fill_hdf5.
 
-            WRITEME
+        Parameters
+        ----------
+        h5file : hdf5 file handle
+            Handle to an hdf5 object.
+        start : int
+            The start index to write data.
+        stop : int
+            The index of the record following the last record to be written.
         """
         ensure_tables()
         # TODO is there any smarter and more efficient way to this?

pylearn2/expr/nnet.py

@@ -114,6 +114,46 @@ def arg_of_softmax(Y_hat):
     return z
 
 
+def arg_of_sigmoid(Y_hat):
+    """
+    Given the output of a call to theano.tensor.nnet.sigmoid,
+    returns the argument to the sigmoid (by tracing the Theano
+    graph).
+
+    Parameters
+    ----------
+    Y_hat : Variable
+        T.nnet.sigmoid(Z)
+
+    Returns
+    -------
+    Z : Variable
+        The variable that was passed to T.nnet.sigmoid to create `Y_hat`.
+        Raises an error if `Y_hat` is not actually the output of a theano
+        sigmoid.
+    """
+    assert hasattr(Y_hat, 'owner')
+    owner = Y_hat.owner
+    assert owner is not None
+    op = owner.op
+    if isinstance(op, Print):
+        assert len(owner.inputs) == 1
+        Y_hat, = owner.inputs
+        owner = Y_hat.owner
+        op = owner.op
+    success = False
+    if isinstance(op, T.Elemwise):
+        if isinstance(op.scalar_op, T.nnet.sigm.ScalarSigmoid):
+            success = True
+    if not success:
+        raise TypeError("Expected Y_hat to be the output of a sigmoid, "
+                        "but it appears to be the output of " + str(op) +
+                        " of type " + str(type(op)))
+    z, = owner.inputs
+    assert z.ndim == 2
+    return z
+
+
 def kl(Y, Y_hat, batch_axis):
     """
     Warning: This function expects a sigmoid nonlinearity in the
@@ -323,4 +363,4 @@ def compute_f1(precision, recall):
     """
     f1 = (2. * precision * recall /
           T.maximum(1, precision + recall))
-    return f1
+    return f1

pylearn2/expr/tests/test_nnet.py

@@ -14,12 +14,13 @@
 from theano import tensor as T
 
 from pylearn2.models.mlp import MLP, Sigmoid
+from pylearn2.expr.nnet import arg_of_sigmoid
 from pylearn2.expr.nnet import pseudoinverse_softmax_numpy
 from pylearn2.expr.nnet import softmax_numpy
 from pylearn2.expr.nnet import softmax_ratio
 from pylearn2.expr.nnet import compute_recall
 from pylearn2.expr.nnet import kl
-from pylearn2.expr.nnet import elemwise_kl 
+from pylearn2.expr.nnet import elemwise_kl
 from pylearn2.utils import sharedX
 
 
@@ -83,7 +84,7 @@ def test_kl():
     """
     init_mode = theano.config.compute_test_value
     theano.config.compute_test_value = 'raise'
-    
+
     try:
         mlp = MLP(layers=[Sigmoid(dim=10, layer_name='Y', irange=0.1)],
                   nvis=10)
@@ -101,7 +102,7 @@ def test_kl():
         np.testing.assert_raises(ValueError, kl, Y, Y_hat, 1)
         Y.tag.test_value[2][3] = -0.1
         np.testing.assert_raises(ValueError, kl, Y, Y_hat, 1)
-    
+
     finally:
         theano.config.compute_test_value = init_mode
 
@@ -112,10 +113,10 @@ def test_elemwise_kl():
     input.
     """
     init_mode = theano.config.compute_test_value
-    theano.config.compute_test_value = 'raise' 
-    
+    theano.config.compute_test_value = 'raise'
+
     try:
-        mlp = MLP(layers=[Sigmoid(dim=10, layer_name='Y', irange=0.1)], 
+        mlp = MLP(layers=[Sigmoid(dim=10, layer_name='Y', irange=0.1)],
                   nvis=10)
         X = mlp.get_input_space().make_theano_batch()
         Y = mlp.get_output_space().make_theano_batch()
@@ -131,8 +132,29 @@ def test_elemwise_kl():
         np.testing.assert_raises(ValueError, elemwise_kl, Y, Y_hat)
         Y.tag.test_value[2][3] = -0.1
         np.testing.assert_raises(ValueError, elemwise_kl, Y, Y_hat)
-    
+
     finally:
         theano.config.compute_test_value = init_mode
 
-   
+def test_arg_of_sigmoid_good():
+    """
+    Tests that arg_of_sigmoid works when given a good input.
+    """
+
+    X = T.matrix()
+    Y = T.nnet.sigmoid(X)
+    Z = arg_of_sigmoid(Y)
+    assert X is Z
+
+def test_arg_of_sigmoid_bad():
+    """
+    Tests that arg_of_sigmoid raises an error when given a bad input.
+    """
+
+    X = T.matrix()
+    Y = T.nnet.softmax(X)
+    try:
+        Z = arg_of_sigmoid(Y)
+    except TypeError:
+        return
+    assert False # Should have failed