very basic GP_regression demo is working

GPy/__init__.py

@@ -6,5 +6,5 @@ import kern
 import models
 import inference
 import util
-import examples
+#import examples
 from core import priors

GPy/core/model.py

@@ -9,7 +9,7 @@ import sys, pdb
 from parameterised import parameterised, truncate_pad
 import priors
 from ..util.linalg import jitchol
-from ..inference import optimization, SGD
+from ..inference import optimization
 
 class model(parameterised):
     def __init__(self):

GPy/examples/GP_regression_demo.py

@@ -19,18 +19,14 @@ pb.close('all')
 X = np.random.uniform(-3.,3.,(20,1))
 Y = np.sin(X)+np.random.randn(20,1)*0.05
 
-# construct kernel
-rbf =  GPy.kern.rbf(1)
-noise = GPy.kern.white(1)
-kernel = rbf + noise
-
 # create simple GP model
-m = GPy.models.GP_regression(X,Y, kernel=kernel)
+m = GPy.models.GP_regression(X,Y)
 
 # contrain all parameters to be positive
 m.constrain_positive('')
+
 # optimize and plot
-m.optimize('rasm', max_f_eval = 1000)
+m.optimize('tnc', max_f_eval = 1000)
 m.plot()
 print(m)
 
@@ -41,11 +37,6 @@ print(m)
 X = np.random.uniform(-3.,3.,(40,2))
 Y = np.sin(X[:,0:1]) * np.sin(X[:,1:2])+np.random.randn(40,1)*0.05
 
-# construct kernel
-rbf =  GPy.kern.rbf(2)
-noise = GPy.kern.white(2)
-kernel = rbf + noise
-
 # create simple GP model
 m = GPy.models.GP_regression(X,Y)
 
@@ -53,7 +44,7 @@ m = GPy.models.GP_regression(X,Y)
 m.constrain_positive('')
 # optimize and plot
 pb.figure()
-m.optimize('rasm', max_f_eval = 1000)
+m.optimize('tnc', max_f_eval = 1000)
 m.plot()
 print(m)
 

GPy/kern/kern.py

@@ -166,8 +166,8 @@ class kern(parameterised):
         slices1, slices2 = self._process_slices(slices1,slices2)
         if X2 is None:
             X2 = X
-        target = np.zeros((X.shape[0],X2.shape[0],self.Nparam))
-        [p.dK_dtheta(X[s1,i_s],X2[s2,i_s],target[s1,s2,ps]) for p,i_s,ps,s1,s2 in zip(self.parts, self.input_slices, self.param_slices, slices1, slices2)]
+        target = np.zeros(self.Nparam)
+        [p.dK_dtheta(partial,X[s1,i_s],X2[s2,i_s],target[ps]) for p,i_s,ps,s1,s2 in zip(self.parts, self.input_slices, self.param_slices, slices1, slices2)]
         return target
 
     def dK_dX(self,X,X2=None,slices1=None,slices2=None):
@@ -185,11 +185,13 @@ class kern(parameterised):
         [p.Kdiag(X[s,i_s],target=target[s]) for p,i_s,s in zip(self.parts,self.input_slices,slices)]
         return target
 
-    def dKdiag_dtheta(self,X,slices=None):
+    def dKdiag_dtheta(self,partial,X,slices=None):
         assert X.shape[1]==self.D
+        assert len(partial.shape)==1
+        assert partial.size==X.shape[0]
         slices = self._process_slices(slices,False)
-        target = np.zeros((X.shape[0],self.Nparam))
-        [p.dKdiag_dtheta(X[s,i_s],target[s,ps]) for p,i_s,s,ps in zip(self.parts,self.input_slices,slices,self.param_slices)]
+        target = np.zeros(self.Nparam)
+        [p.dKdiag_dtheta(partial,X[s,i_s],target[ps]) for p,i_s,s,ps in zip(self.parts,self.input_slices,slices,self.param_slices)]
         return target
 
     def dKdiag_dX(self, X, slices=None):

GPy/kern/rbf.py

@@ -50,14 +50,13 @@ class rbf(kernpart):
     def Kdiag(self,X,target):
         np.add(target,self.variance,target)
 
-    def dK_dtheta(self,X,X2,target):
-        """Return shape is NxMx(Ntheta)"""
+    def dK_dtheta(self,partial,X,X2,target):
         self._K_computations(X,X2)
-        target[:,:,0] += self._K_dvar
-        target[:,:,1] += self._K_dvar*self.variance*self._K_dist2/self.lengthscale
+        target[0] += np.sum(self._K_dvar*partial)
+        target[1] += np.sum(self._K_dvar*self.variance*self._K_dist2/self.lengthscale*partial)
 
     def dKdiag_dtheta(self,X,target):
-        np.add(target[:,0],1.,target[:,0])
+        target[0] += partial
 
     def dK_dX(self,X,X2,target):
         self._K_computations(X,X2)

GPy/kern/white.py

@@ -37,10 +37,10 @@ class white(kernpart):
     def Kdiag(self,X,target):
         target += self.variance
 
-    def dK_dtheta(self,X,X2,target):
+    def dK_dtheta(self,partial,X,X2,target):
         if X.shape==X2.shape:
             if np.all(X==X2):
-                np.add(target[:,:,0],np.eye(X.shape[0]),target[:,:,0])
+                target += np.trace(partial)
 
     def dKdiag_dtheta(self,X,target):
         np.add(target[:,0],1.,target[:,0])
@@ -50,7 +50,7 @@ class white(kernpart):
 
     def dKdiag_dX(self,X,target):
         pass
-        
+
     def psi0(self,Z,mu,S,target):
         target += self.variance
 

GPy/models/GP_EP.py

@@ -7,7 +7,7 @@ import pylab as pb
 from scipy import stats, linalg
 from .. import kern
 from ..inference.Expectation_Propagation import EP,Full
-from ..inference.likelihoods import likelihood,probit,poisson,gaussian
+from ..inference.likelihoods import likelihood,probit#,poisson,gaussian
 from ..core import model
 from ..util.linalg import pdinv,jitchol
 from ..util.plot import gpplot

GPy/models/GP_regression.py

@@ -29,7 +29,7 @@ class GP_regression(model):
 
     def __init__(self,X,Y,kernel=None,normalize_X=False,normalize_Y=False, Xslices=None):
         if kernel is None:
-            kernel = kern.rbf(X.shape[1]) + kern.bias(X.shape[1]) + kern.white(X.shape[1])
+            kernel = kern.rbf(X.shape[1]) + kern.white(X.shape[1])
 
         # parse arguments
         self.Xslices = Xslices
@@ -103,7 +103,7 @@ class GP_regression(model):
         return dL_dK
 
     def log_likelihood_gradients(self):
-        return self.kern.dK_dtheta(self.X,partial=self.dL_dK())
+        return self.kern.dK_dtheta(partial=self.dL_dK(),X=self.X)
 
     def predict(self,Xnew, slices=None):
         """

GPy/models/__init__.py

@@ -6,6 +6,6 @@ from GP_regression import GP_regression
 from sparse_GP_regression import sparse_GP_regression
 from GPLVM import GPLVM
 from warped_GP import warpedGP
-from simple_GP_EP import GP_EP
+from GP_EP import GP_EP
 from generalized_FITC import generalized_FITC
 from sparse_GPLVM import sparse_GPLVM

GPy/models/generalized_FITC.py

@@ -9,7 +9,7 @@ from .. import kern
 from ..core import model
 from ..util.linalg import pdinv,mdot
 from ..util.plot import gpplot
-from ..inference.Expectation_Propagation import EP,Full,DTC,FITC
+from ..inference.Expectation_Propagation import EP,Full,FITC
 from ..inference.likelihoods import likelihood,probit
 
 class generalized_FITC(model):
@@ -17,14 +17,12 @@ class generalized_FITC(model):
         """
         Naish-Guzman, A. and Holden, S. (2008) implemantation of EP with FITC.
 
-        Arguments
-        ---------
-        X : input observations
-        likelihood : Output's likelihood (likelihood class)
-        kernel : a GPy kernel
-        inducing : Either an array specifying the inducing points location or a scalar defining their number.
-        epsilon_ep : EP convergence criterion, maximum squared difference allowed between mean updates to stop iterations (float)
-        powerep : Power-EP parameters (eta,delta) - 2x1 numpy array (floats)
+        :param X: input observations
+        :param likelihood: Output's likelihood (likelihood class)
+        :param kernel: a GPy kernel
+        :param inducing:  Either an array specifying the inducing points location or a scalar defining their number.
+        :param epsilon_ep: EP convergence criterion, maximum squared difference allowed between mean updates to stop iterations (float)
+        :param powerep: Power-EP parameters (eta,delta) - 2x1 numpy array (floats)
         """
         assert isinstance(kernel,kern.kern)
         self.likelihood = likelihood