From 06dd27c634a87fe86596bf09790061dbf6255fb9 Mon Sep 17 00:00:00 2001
From: James Hensman <james.hensman@gmail.com>
Date: Mon, 24 Feb 2014 14:49:20 +0000
Subject: [PATCH] input senitivity in stationary

---
 GPy/kern/_src/stationary.py | 54 ++++---------------------------------
 1 file changed, 5 insertions(+), 49 deletions(-)

diff --git a/GPy/kern/_src/stationary.py b/GPy/kern/_src/stationary.py
index dde26e63..e8586d07 100644
--- a/GPy/kern/_src/stationary.py
+++ b/GPy/kern/_src/stationary.py
@@ -35,6 +35,9 @@ class Stationary(Kern):
     def K_of_r(self, r):
         raise NotImplementedError, "implement the covaraiance functino and a fn of r to use this class"
 
+    def dK_dr(self, r):
+        raise NotImplementedError, "implement the covaraiance functino and a fn of r to use this class"
+
     def K(self, X, X2=None):
         r = self._scaled_dist(X, X2)
         return self.K_of_r(r)
@@ -92,55 +95,8 @@ class Stationary(Kern):
     def gradients_X_diag(self, dL_dKdiag, X):
         return np.zeros(X.shape)
 
-    def add(self, other, tensor=False):
-        if not tensor:
-            return StatAdd(self, other)
-        else:
-            return super(Stationary, self).add(other, tensor)
-
-    def prod(self, other, tensor=False):
-        if not tensor:
-            return StatProd(self, other)
-        else:
-            return super(Stationary, self).prod(other, tensor)
-
-
-
-class StatAdd(Stationary):
-    """
-    Addition of two Stationary kernels on the same space is still stationary.
-
-    If you need to add two (stationary) kernels on separate spaces, use the generic add class.
-    """
-    def __init__(self, k1, k2):
-        assert isinstance(k1, Stationary)
-        assert isinstance(k2, Stationary)
-        self.k1, self.k2 = k1, k2
-        self.add_parameters(k1, k2)
-
-    def K_of_r(self, r):
-        return self.k1.K(r) + self.k2.K(r)
-
-    def dK_dr(self, r):
-        return self.k1.dK_dr + self.k2.dK_dr(r)
-
-class StatProd(Stationary):
-    """
-    Product of two Stationary kernels on the same space is still stationary.
-
-    If you need to multiply two (stationary) kernels on separate spaces, use the generic Prod class.
-    """
-    def __init__(self, k1, k2):
-        assert isinstance(k1, Stationary)
-        assert isinstance(k2, Stationary)
-        self.k1, self.k2 = k1, k2
-        self.add_parameters(k1, k2)
-
-    def K_of_r(self, r):
-        return self.k1.K(r) * self.k2.K(r)
-
-    def dK_dr(self, r):
-        return self.k1.dK_dr(r) * self.k2.K_of_r(r) + self.k2.dK_dr(r) * self.k1.K_of_r(r)
+    def input_sensitivity(self):
+        return np.ones(self.input_dim)/self.lengthscale
 
 class Exponential(Stationary):
     def __init__(self, input_dim, variance=1., lengthscale=None, ARD=False, name='Exponential'):