diff --git a/GPy/kern/constructors.py b/GPy/kern/constructors.py
index 392f43ba..b60c7479 100644
--- a/GPy/kern/constructors.py
+++ b/GPy/kern/constructors.py
@@ -450,6 +450,18 @@ def prod(k1,k2,tensor=False):
 def symmetric(k):
     """
     Construct a symmetric kernel from an existing kernel
+
+    The symmetric kernel works by adding two GP functions together, and computing the overall covariance.
+
+    Let f ~ GP(x | 0, k(x, x')). Now let g = f(x) + f(-x).
+
+    It's easy to see that g is a symmetric function: g(x) = g(-x).
+
+    by construction, g, is a gaussian Process with mean 0 and covariance
+
+    k(x, x') + k(-x, x') + k(x, -x') + k(-x, -x')
+
+    This constructor builds a covariance function of this form from the initial kernel
     """
     k_ = k.copy()
     k_.parts = [symmetric.Symmetric(p) for p in k.parts]
diff --git a/GPy/kern/parts/prod.py b/GPy/kern/parts/prod.py
index 7441ae9f..f517262c 100644
--- a/GPy/kern/parts/prod.py
+++ b/GPy/kern/parts/prod.py
@@ -133,12 +133,12 @@ class Prod(Kernpart):
                 self.k1.K(X[:,self.slice1],X2[:,self.slice1],self._K1)
                 self.k2.K(X[:,self.slice2],X2[:,self.slice2],self._K2)
 
-    def __getstate__(self):
-        return [self.k1, self.k2, self.slice1, self.slice2, self.name, self.input_dim, self.num_params]
+    #def __getstate__(self):
+        #return [self.k1, self.k2, self.slice1, self.slice2, self.name, self.input_dim, self.num_params]
 
-    def __setstate__(self, state):
-        self.k1, self.k2, self.slice1, self.slice2, self.name, self.input_dim, self.num_params = state
-        self._X, self._X2, self._params = np.empty(shape=(3,1))
+    #def __setstate__(self, state):
+        #self.k1, self.k2, self.slice1, self.slice2, self.name, self.input_dim, self.num_params = state
+        #self._X, self._X2, self._params = np.empty(shape=(3,1))