Merge pull request #318 from goodfeli/CAReduce_R_op

theano/gof/op.py

@@ -449,6 +449,38 @@ class PureOp(object):
     # Python implementation #
     #########################
 
+
+    def R_op(self, inputs, eval_points):
+        """
+
+        This method is primarily used by tensor.Rop
+
+        Suppose the op outputs
+
+        [ f_1(inputs), ..., f_n(inputs) ]
+
+        inputs: a Variable or list of Variables
+        eval_points: a Variable or list of Variables with
+                    the same length as inputs. Each element
+                    of eval_points specifies the value of
+                    the corresponding input at the point
+                    where the R op is to be evaluated.
+
+
+        returns: a list of n elements
+                    rval[i] should be Rop(f = f_i(inputs),
+                                          wrt = inputs,
+                                          eval_points = eval_points)
+
+        """
+
+        raise NotImplementedError(str(self)+' of type '+str(self.__class__.__name__)
+                +" does not "
+                "implement R_op. If this is a theano op, write to the "
+                "theano-dev mailing list for assistance. If it is your "
+                "own op, implement the R_op method.")
+
+
     def perform(self, node, inputs, output_storage):
         """
         Required:  Calculate the function on the inputs and put the variables in the

theano/tensor/elemwise.py

@@ -1007,6 +1007,7 @@ class CAReduce(Op):
         if scalar_op.nin not in [-1, 2] or scalar_op.nout != 1:
             raise NotImplementedError("CAReduce only supports binary functions with a single output.")
         self.scalar_op = scalar_op
+
         if axis is None:
             self.axis = axis
         elif isinstance(axis, int):

theano/tensor/tensor_grad.py

@@ -65,8 +65,10 @@ def Rop(f, wrt, eval_points):
 
     assert len(wrt) == len(eval_points)
 
+    #check that each element of wrt corresponds to an element
+    #of eval_points with the same dimensionality
     for pack in enumerate(zip(wrt, eval_points)):
-        i = pack[0]
+        i =  pack[0]
         wrt_elem, eval_point = pack[1]
 
         wrt_elem = as_tensor_variable(wrt_elem)
@@ -82,15 +84,13 @@ def Rop(f, wrt, eval_points):
     seen_nodes = {}
 
     def _traverse(node):
+        """ TODO: writeme """
         if node is None:
             return None
         else:
             op = node.op
             inputs = node.inputs
-            if not hasattr(op, 'R_op'):
-                raise Exception((' R_op was not implemented for %s'
-                                      ' operation. Email the mailing list'
-                                      ' for help') % op.__class__.__name__)
+
             # Compute the evaluation points corresponding to each of the
             # inputs of the node
             local_eval_points = []