Merge pull request #4263 from shabanian/opt

theano/tensor/opt.py

@@ -3539,6 +3539,39 @@ def local_join_make_vector(node):
         return [ret]
 
 
+#################
+#  speed/memory #
+#################
+@register_canonicalize
+@register_specialize
+@gof.local_optimizer([T.elemwise.Sum])
+def local_sumsqr2dot(node):
+    """
+    This optimization detects T.sqr( W.dimshuffle('x',0,1) * G.dimshuffle(0,'x',1) ).sum(axis=(1,2))
+     and converts this to T.dot(T.sqr(G), T.sqr(W).sum(axis=0)).
+    """
+    if (isinstance(node.op, T.elemwise.Sum) and
+            isinstance(node.op.scalar_op, theano.scalar.basic.Add) and node.op.axis == (1, 2)):
+        in1 = node.inputs[0]
+        out = node.outputs[0]
+
+        if (in1.owner and isinstance(in1.owner.op, T.Elemwise) and isinstance(in1.owner.op.scalar_op, theano.scalar.basic.Sqr)):
+            in_sqr = in1.owner.inputs[0]
+            if (in_sqr.owner and isinstance(in_sqr.owner.op, T.Elemwise) and
+                    isinstance(in_sqr.owner.op.scalar_op, theano.scalar.basic.Mul) and len(in_sqr.owner.inputs) == 2):
+                in_mul1, in_mul2 = in_sqr.owner.inputs
+
+                if (isinstance(in_mul1.owner.op, T.elemwise.DimShuffle) and in_mul1.owner.op.new_order == ('x', 0, 1) and
+                        isinstance(in_mul2.owner.op, T.elemwise.DimShuffle) and in_mul2.owner.op.new_order == (0, 'x', 1)):
+                    W = in_mul1.owner.inputs[0]
+                    G = in_mul2.owner.inputs[0]
+
+                    new_out = T.dot(T.sqr(G), T.sqr(W).sum(axis=0))
+                    if new_out.dtype != out.dtype:
+                        new_out = T.cast(new_out, dtype=out.dtype)
+                    return [new_out]
+
+
 #################
 # Exp stability #
 #################

theano/tensor/tests/test_opt.py

@@ -6173,6 +6173,27 @@ def test_local_zero_div():
             assert theano.tensor.get_scalar_constant_value(output) == 0
 
 
+def test_local_sumsqr2dot():
+    G = matrix('G')
+    W = matrix('W')
+
+    y = T.sqr(W.dimshuffle('x', 0, 1) * G.dimshuffle(0, 'x', 1)).sum(axis=(1, 2))
+    MODE = theano.compile.get_default_mode().including('local_sumsqr2dot')
+
+    f = function([W, G], y, mode=MODE)
+
+    w_val = numpy.random.rand(4, 3).astype(config.floatX)
+    g_val = numpy.random.rand(5, 3).astype(config.floatX)
+
+    f_val = f(w_val, g_val)
+    f_test = numpy.dot(numpy.square(g_val), numpy.square(w_val).sum(axis=0))
+
+    assert numpy.allclose(f_val, f_test)
+    assert any(isinstance(n.op, (tensor.basic.Dot, tensor.blas.Dot22,
+                                 tensor.blas.Gemv, tensor.blas_c.CGemv))
+               for n in f.maker.fgraph.toposort())
+
+
 def test_local_expm1():
     x = matrix('x')
     u = T.scalar('u')