Fixed the computation of Lop

The formula before was wrong, it does not work when inputs are matrices, because it depends on which axis y is computed. I believe this formulation is much better.

Fixed the computation of Lop
2d85bb2a · Razvan Pascanu · daef1d0c · 2d85bb2a
--- a/theano/tensor/tests/test_rop.py
+++ b/theano/tensor/tests/test_rop.py
@@ -130,9 +130,8 @@ class test_RopLop(unittest.TestCase):
        vv = numpy.asarray(self.rng.uniform(size=out_shape), theano.config.floatX)
        yv = TT.Lop(y, self.mx, self.v)
        lop_f = function([self.mx, self.v], yv)
-        sy, _ = theano.scan( lambda i,y,x,v: (TT.grad(y[i]*v[i],x))[i],
-                           sequences = TT.arange(y.shape[0]),
+        sy = TT.grad((self.v*y).sum(), self.mx)
-                           non_sequences = [y,self.mx,self.v])
        scan_f = function([self.mx, self.v], sy)