added Brownian motion

GPy/kern/__init__.py

@@ -2,8 +2,8 @@ from _src.rbf import RBF
 from _src.white import White
 from _src.kern import Kern
 from _src.linear import Linear
-#import bias
+from _src.brownian import Brownian
-#import Brownian
+#from _src.bias import Bias
 #import coregionalize
 #import exponential
 #import eq_ode1

GPy/kern/_src/Brownian.py

@@ -1,65 +0,0 @@
-# Copyright (c) 2012, GPy authors (see AUTHORS.txt).
-# Licensed under the BSD 3-clause license (see LICENSE.txt)
-
-
-from kernpart import Kernpart
-import numpy as np
-
-def theta(x):
-    """Heavisdie step function"""
-    return np.where(x>=0.,1.,0.)
-
-class Brownian(Kernpart):
-    """
-    Brownian Motion kernel.
-
-    :param input_dim: the number of input dimensions
-    :type input_dim: int
-    :param variance:
-    :type variance: float
-    """
-    def __init__(self,input_dim,variance=1.):
-        self.input_dim = input_dim
-        assert self.input_dim==1, "Brownian motion in 1D only"
-        self.num_params = 1
-        self.name = 'Brownian'
-        self._set_params(np.array([variance]).flatten())
-
-    def _get_params(self):
-        return self.variance
-
-    def _set_params(self,x):
-        assert x.shape==(1,)
-        self.variance = x
-
-    def _get_param_names(self):
-        return ['variance']
-
-    def K(self,X,X2,target):
-        if X2 is None:
-            X2 = X
-        target += self.variance*np.fmin(X,X2.T)
-
-    def Kdiag(self,X,target):
-        target += self.variance*X.flatten()
-
-    def _param_grad_helper(self,dL_dK,X,X2,target):
-        if X2 is None:
-            X2 = X
-        target += np.sum(np.fmin(X,X2.T)*dL_dK)
-
-    def dKdiag_dtheta(self,dL_dKdiag,X,target):
-        target += np.dot(X.flatten(), dL_dKdiag)
-
-    def gradients_X(self,dL_dK,X,X2,target):
-        raise NotImplementedError, "TODO"
-        #target += self.variance
-        #target -= self.variance*theta(X-X2.T)
-        #if X.shape==X2.shape:
-            #if np.all(X==X2):
-                #np.add(target[:,:,0],self.variance*np.diag(X2.flatten()-X.flatten()),target[:,:,0])
-
-
-    def dKdiag_dX(self,dL_dKdiag,X,target):
-        target += self.variance*dL_dKdiag[:,None]
-

GPy/kern/_src/brownian.py

@@ -0,0 +1,50 @@
+# Copyright (c) 2012, GPy authors (see AUTHORS.txt).
+# Licensed under the BSD 3-clause license (see LICENSE.txt)
+
+from kern import Kern
+from ...core.parameterization import Param
+from ...core.parameterization.transformations import Logexp
+import numpy as np
+
+class Brownian(Kern):
+    """
+    Brownian motion in 1D only.
+
+    Negative times are treated as a separate (backwards!) Brownian motion.
+
+    :param input_dim: the number of input dimensions
+    :type input_dim: int
+    :param variance:
+    :type variance: float
+    """
+    def __init__(self, input_dim=1, variance=1., name='Brownian'):
+        assert input_dim==1, "Brownian motion in 1D only"
+        super(Brownian, self).__init__(input_dim, name)
+
+        self.variance = Param('variance', variance, Logexp())
+        self.add_parameters(self.variance)
+
+    def K(self,X,X2=None):
+        if X2 is None:
+            X2 = X
+        return self.variance*np.where(np.sign(X)==np.sign(X2.T),np.fmin(np.abs(X),np.abs(X2.T)), 0.)
+
+    def Kdiag(self,X):
+        return self.variance*np.abs(X.flatten())
+
+    def update_gradients_full(self, dL_dK, X, X2=None):
+        if X2 is None:
+            X2 = X
+        self.variance.gradient = np.sum(dL_dK * np.where(np.sign(X)==np.sign(X2.T),np.fmin(np.abs(X),np.abs(X2.T)), 0.))
+
+    #def update_gradients_diag(self, dL_dKdiag, X):
+        #self.variance.gradient = np.dot(np.abs(X.flatten()), dL_dKdiag)
+
+    #def gradients_X(self, dL_dK, X, X2=None):
+        #if X2 is None:
+            #return np.sum(self.variance*dL_dK*np.abs(X),1)[:,None]
+        #else:
+            #return np.sum(np.where(np.logical_and(np.abs(X)<np.abs(X2.T), np.sign(X)==np.sign(X2)), self.variance*dL_dK,0.),1)[:,None]
+
+
+

GPy/kern/_src/kern.py

@@ -40,8 +40,15 @@ class Kern(Parameterized):
         """Set the gradients of all parameters when doing full (N) inference."""
         raise NotImplementedError
     def update_gradients_sparse(self, dL_dKmm, dL_dKnm, dL_dKdiag, X, Z):
-        """Set the gradients of all parameters when doing sparse (M) inference."""
+        target = np.zeros(self.size)
-        raise NotImplementedError
+        self.update_gradients_diag(dL_dKdiag, X)
+        self._collect_gradient(target)
+        self.update_gradients_full(dL_dKnm, X, Z)
+        self._collect_gradient(target)
+        self.update_gradients_full(dL_dKmm, Z, None)
+        self._collect_gradient(target)
+        self._set_gradient(target)
+
     def update_gradients_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, mu, S, Z):
         """Set the gradients of all parameters when doing variational (M) inference with uncertain inputs."""
         raise NotImplementedError

GPy/kern/_src/linear.py

@@ -74,16 +74,6 @@ class Linear(Kern):
     def Kdiag(self, X):
         return np.sum(self.variances * np.square(X), -1)
 
-    def update_gradients_sparse(self, dL_dKmm, dL_dKnm, dL_dKdiag, X, Z):
-        target = np.zeros(self.size)
-        self.update_gradients_diag(dL_dKdiag, X)
-        self._collect_gradient(target)
-        self.update_gradients_full(dL_dKnm, X, Z)
-        self._collect_gradient(target)
-        self.update_gradients_full(dL_dKmm, Z, None)
-        self._collect_gradient(target)
-        self._set_gradient(target)
-
     def update_gradients_full(self, dL_dK, X, X2=None):
         if self.ARD:
             if X2 is None:

GPy/plotting/matplot_dep/models_plots.py

@@ -92,11 +92,11 @@ def plot_fit(model, plot_limits=None, which_data_rows='all',
                 ax.plot(Xnew, yi[:,None], Tango.colorsHex['darkBlue'], linewidth=0.25)
                 #ax.plot(Xnew, yi[:,None], marker='x', linestyle='--',color=Tango.colorsHex['darkBlue']) #TODO apply this line for discrete outputs.
 
-        
+
         #add error bars for uncertain (if input uncertainty is being modelled)
-        if hasattr(model,"has_uncertain_inputs"):
+        if hasattr(model,"has_uncertain_inputs") and model.has_uncertain_inputs():
-            ax.errorbar(model.X[which_data, free_dims], model.likelihood.data[which_data, 0],
+            ax.errorbar(model.X[which_data_rows, free_dims], model.Y[which_data_rows, which_data_ycols],
-                        xerr=2 * np.sqrt(model.X_variance[which_data, free_dims]),
+                        xerr=2 * np.sqrt(model.X_variance[which_data_rows, free_dims]),
                         ecolor='k', fmt=None, elinewidth=.5, alpha=.5)