Fix description of exponentiated quadratic covariance function and add kernels. (#872)

* Fix description of ExpQuad covariance.

* Add sinc kernel.

* Add ExpQuadCosine covariance

GPy/kern/__init__.py

@@ -38,7 +38,7 @@ from .src.rbf import RBF
 from .src.linear import Linear, LinearFull
 from .src.static import Bias, White, Fixed, WhiteHeteroscedastic, Precomputed
 from .src.brownian import Brownian
-from .src.stationary import Exponential, OU, Matern32, Matern52, ExpQuad, RatQuad, Cosine
+from .src.stationary import Exponential, OU, Matern32, Matern52, ExpQuad, RatQuad, Cosine, Sinc, ExpQuadCosine
 from .src.mlp import MLP
 from .src.periodic import PeriodicExponential, PeriodicMatern32, PeriodicMatern52
 from .src.standard_periodic import StdPeriodic

GPy/kern/src/stationary.py

@@ -621,10 +621,10 @@ class ExpQuad(Stationary):
 
     .. math::
 
-       k(r) = \sigma^2 (1 + \sqrt{5} r + \\frac53 r^2) \exp(- \sqrt{5} r)
+       k(r) = \sigma^2 \exp(- 0.5 r^2)
 
     notes::
-     - Yes, this is exactly the same as the RBF covariance function, but the
+     - This is exactly the same as the RBF covariance function, but the
        RBF implementation also has some features for doing variational kernels
        (the psi-statistics).
 
@@ -657,6 +657,14 @@ class ExpQuad(Stationary):
         return -r*self.K_of_r(r)
 
 class Cosine(Stationary):
+    """
+    Cosine Covariance function
+    
+    .. math::
+
+        k(r) = \sigma^2 \cos(r)
+
+    """
     def __init__(self, input_dim, variance=1., lengthscale=None, ARD=False, active_dims=None, name='Cosine'):
         super(Cosine, self).__init__(input_dim, variance, lengthscale, ARD, active_dims, name)
 
@@ -666,6 +674,62 @@ class Cosine(Stationary):
     def dK_dr(self, r):
         return -self.variance * np.sin(r)
 
+class ExpQuadCosine(Stationary):
+    """
+    Exponentiated quadratic multiplied by cosine covariance function (spectral mixture kernel).
+    
+    .. math::
+
+        k(r) = \sigma^2 \exp(-2\pi^2r^2)\cos(2\pi r/T)
+
+    """
+    
+    def __init__(self, input_dim, variance=1., lengthscale=None, period=1., ARD=False, active_dims=None, name='ExpQuadCosine'):
+        super(ExpQuadCosine, self).__init__(input_dim, variance, lengthscale, ARD, active_dims, name)
+        self.period = Param('period', period, Logexp())
+        self.link_parameters(self.period)
+        
+    def K_of_r(self, r):
+        return self.variance * np.exp(-2*np.pi**2*r**2)*np.cos(2*np.pi*r/self.period)
+
+    def dK_dr(self, r):
+        return -4*np.pi**2*r*self.K_of_r(r) - self.variance * 2*np.pi/self.period*np.exp(-2*np.pi**2*r**2)*np.sin(2*np.pi*r/self.period)
+
+    def update_gradients_full(self, dL_dK, X, X2=None):
+        super(ExpQuadCosine, self).update_gradients_full(dL_dK, X, X2)
+        r = self._scaled_dist(X, X2)
+        r2 = np.square(r)
+        dK_dperiod = self.variance * 2*np.pi*r/self.period**2*np.exp(-2*np.pi**2*r**2)*np.sin(2*np.pi*r/self.period)
+        grad = np.sum(dL_dK*dK_dperiod)
+        self.period.gradient = grad
+
+    def update_gradients_diag(self, dL_dKdiag, X):
+        super(ExpQuadCosine, self).update_gradients_diag(dL_dKdiag, X)
+        self.period.gradient = 0.
+
+    
+    
+class Sinc(Stationary):
+    """
+    Sinc Covariance function
+    
+    .. math::
+
+        k(r) = \sigma^2 \sinc(\pi r)
+
+    """
+    
+    def __init__(self, input_dim, variance=1., lengthscale=None, ARD=False, active_dims=None, name='Sinc'):
+        super(Sinc, self).__init__(input_dim, variance, lengthscale, ARD, active_dims, name)
+
+    def K_of_r(self, r):
+        return self.variance * np.sinc(2*r)
+
+    def dK_dr(self, r):
+        # small angle approximation to avoid divide by zero errors.
+        return np.where(r<1e-5, -self.variance*4/3*np.pi*np.pi*r, self.variance/r * (np.cos(2*np.pi*r)-np.sinc(2*r)))
+
+    
 
 class RatQuad(Stationary):
     """
@@ -677,7 +741,6 @@ class RatQuad(Stationary):
 
     """
 
-
     def __init__(self, input_dim, variance=1., lengthscale=None, power=2., ARD=False, active_dims=None, name='RatQuad'):
         super(RatQuad, self).__init__(input_dim, variance, lengthscale, ARD, active_dims, name)
         self.power = Param('power', power, Logexp())

GPy/testing/kernel_tests.py

@@ -426,6 +426,22 @@ class KernelGradientTestsContinuous(unittest.TestCase):
         k.randomize()
         self.assertTrue(check_kernel_gradient_functions(k, X=self.X, X2=self.X2, verbose=verbose))
 
+    def test_Cosine(self):
+        # Don't test Cosine directly as it fails positive definite test.
+        k = GPy.kern.RBF(self.D-1, ARD=False)*GPy.kern.Cosine(self.D-1, ARD=True)
+        k.randomize()
+        self.assertTrue(check_kernel_gradient_functions(k, X=self.X, X2=self.X2, verbose=verbose))
+
+    def test_ExpQuadCosine(self):
+        k = GPy.kern.ExpQuadCosine(self.D-1, ARD=True)
+        k.randomize()
+        self.assertTrue(check_kernel_gradient_functions(k, X=self.X, X2=self.X2, verbose=verbose))
+
+    def test_Sinc(self):
+        k = GPy.kern.Sinc(self.D-1, ARD=True)
+        k.randomize()
+        self.assertTrue(check_kernel_gradient_functions(k, X=self.X, X2=self.X2, verbose=verbose))
+
     def test_RatQuad(self):
         k = GPy.kern.RatQuad(self.D-1, ARD=True)
         k.randomize()

GPy/testing/state_space_main_tests.py

@@ -722,9 +722,9 @@ class StateSpaceKernelsTests(np.testing.TestCase):
             # plotting <-
         # 2D measurement, 3 ts_no <-
             
-    def test_continuos_ss(self,plot=False):
+    def test_continuous_ss(self,plot=False):
         """
-        This function tests the continuos state-space model.
+        This function tests the continuous state-space model.
         """                    
                 
         # 1D measurements, 1 ts_no ->
@@ -974,4 +974,4 @@ if __name__ == '__main__':
     #res = tt.test_discrete_ss_2D(plot=False) 
     #res = tt.test_continuos_ss(plot=True)
     
- 
+