Add check for function output

theano/scan_module/tests/test_scan.py

@@ -2436,10 +2436,10 @@ class T_Scan(unittest.TestCase):
         # recurrent output of a scan node associated with a state with a
         # state with broadcastable dimensions
 
-        x = tensor.col()
-        seq = tensor.col()
+        x = tensor.dcol()
+        seq = tensor.dcol()
         outputs_info=[x, tensor.zeros_like(x)]
-        (out1, out2), updates = theano.scan(lambda a, b, c : (a + b, b + c),
+        (out1, out2), updates = theano.scan(lambda a, b, c : (a + b, a + c),
                                             sequences=seq,
                                             outputs_info=outputs_info)
 
@@ -2452,6 +2452,22 @@ class T_Scan(unittest.TestCase):
         fct = theano.function([x, seq],
                               [out1_direct, out2_direct])
 
+        # Test that the function returns valid outputs
+        x_val = numpy.arange(0, 4)[:, None]
+        seq_val = numpy.arange(4, 8)[:, None]
+
+        out1, out2 = fct(x_val, seq_val)
+
+        expected_out1 = numpy.zeros((5, 4, 1))
+        expected_out2 = numpy.zeros((5, 4, 1))
+        for i in range(4):
+            expected_out2[i + 1] = expected_out2[i] + seq_val[i]
+        for i in range(5):
+            expected_out1[i] = expected_out2[i] + x_val
+
+        utt.assert_allclose(out1, expected_out1)
+        utt.assert_allclose(out2, expected_out2)
+
     def test_infer_shape(self):
         # Test for a crash in scan.infer_shape when using both
         # an until condition and random sampling in the inner function.