renaming: posterior_variationa -> variational_posterior

GPy/examples/regression.py

@@ -16,7 +16,7 @@ def olympic_marathon_men(optimize=True, plot=True):
     m = GPy.models.GPRegression(data['X'], data['Y'])
 
     # set the lengthscale to be something sensible (defaults to 1)
-    m['rbf_lengthscale'] = 10
+    m.kern.lengthscale = 10.
 
     if optimize:
         m.optimize('bfgs', max_iters=200)
@@ -41,11 +41,10 @@ def coregionalization_toy2(optimize=True, plot=True):
     Y = np.vstack((Y1, Y2))
 
     #build the kernel
-    k1 = GPy.kern.RBF(1) + GPy.kern.bias(1)
+    k1 = GPy.kern.RBF(1) + GPy.kern.Bias(1)
-    k2 = GPy.kern.coregionalize(2,1)
+    k2 = GPy.kern.Coregionalize(2,1)
     k = k1**k2
     m = GPy.models.GPRegression(X, Y, kernel=k)
-    m.constrain_fixed('.*rbf_var', 1.)
 
     if optimize:
         m.optimize('bfgs', max_iters=100)
@@ -86,11 +85,13 @@ def coregionalization_sparse(optimize=True, plot=True):
     """
     #fetch the data from the non sparse examples
     m = coregionalization_toy2(optimize=False, plot=False)
-    X, Y = m.X, m.likelihood.Y
+    X, Y = m.X, m.Y
+
+    k = GPy.kern.RBF(1)**GPy.kern.Coregionalize(2)
 
     #construct a model
-    m = GPy.models.SparseGPRegression(X,Y)
+    m = GPy.models.SparseGPRegression(X,Y, num_inducing=25, kernel=k)
-    m.constrain_fixed('iip_\d+_1') # don't optimize the inducing input indexes
+    m.Z[:,1].fix() # don't optimize the inducing input indexes
 
     if optimize:
         m.optimize('bfgs', max_iters=100, messages=1)
@@ -128,7 +129,7 @@ def epomeo_gpx(max_iters=200, optimize=True, plot=True):
                    np.random.randint(0, 4, num_inducing)[:, None]))
 
     k1 = GPy.kern.RBF(1)
-    k2 = GPy.kern.coregionalize(output_dim=5, rank=5)
+    k2 = GPy.kern.Coregionalize(output_dim=5, rank=5)
     k = k1**k2
 
     m = GPy.models.SparseGPRegression(t, Y, kernel=k, Z=Z, normalize_Y=True)
@@ -322,7 +323,7 @@ def toy_ARD(max_iters=1000, kernel_type='linear', num_samples=300, D=4, optimize
         kernel = GPy.kern.RBF_inv(X.shape[1], ARD=1)
     else:
         kernel = GPy.kern.RBF(X.shape[1], ARD=1)
-    kernel += GPy.kern.White(X.shape[1]) + GPy.kern.bias(X.shape[1])
+    kernel += GPy.kern.White(X.shape[1]) + GPy.kern.Bias(X.shape[1])
     m = GPy.models.GPRegression(X, Y, kernel)
     # len_prior = GPy.priors.inverse_gamma(1,18) # 1, 25
     # m.set_prior('.*lengthscale',len_prior)
@@ -361,7 +362,7 @@ def toy_ARD_sparse(max_iters=1000, kernel_type='linear', num_samples=300, D=4, o
         kernel = GPy.kern.RBF_inv(X.shape[1], ARD=1)
     else:
         kernel = GPy.kern.RBF(X.shape[1], ARD=1)
-    #kernel += GPy.kern.bias(X.shape[1])
+    #kernel += GPy.kern.Bias(X.shape[1])
     X_variance = np.ones(X.shape) * 0.5
     m = GPy.models.SparseGPRegression(X, Y, kernel, X_variance=X_variance)
     # len_prior = GPy.priors.inverse_gamma(1,18) # 1, 25

GPy/kern/_src/add.py

@@ -45,9 +45,6 @@ class Add(Kern):
     def update_gradients_full(self, dL_dK, X):
         [p.update_gradients_full(dL_dK, X[:,i_s]) for p, i_s in zip(self._parameters_, self.input_slices)]
 
-    def update_gradients_sparse(self, dL_dKmm, dL_dKnm, dL_dKdiag, X, Z):
-        [p.update_gradients_sparse(dL_dKmm, dL_dKnm, dL_dKdiag, X[:,i_s], Z[:,i_s]) for p, i_s in zip(self._parameters_, self.input_slices)]
-
     def gradients_X(self, dL_dK, X, X2=None):
         """Compute the gradient of the objective function with respect to X.
 

GPy/kern/_src/coregionalize.py

@@ -129,7 +129,7 @@ class Coregionalize(Kern):
 
     def update_gradients_diag(self, dL_dKdiag, X):
         index = np.asarray(X, dtype=np.int).flatten()
-        dL_dKdiag_small = np.array([dL_dKdiag[index==i] for i in xrange(output_dim)])
+        dL_dKdiag_small = np.array([dL_dKdiag[index==i].sum() for i in xrange(self.output_dim)])
         self.W.gradient = 2.*self.W*dL_dKdiag_small[:, None]
         self.kappa.gradient = dL_dKdiag_small
 

GPy/kern/_src/kern.py

@@ -26,11 +26,11 @@ class Kern(Parameterized):
         raise NotImplementedError
     def Kdiag(self, Xa):
         raise NotImplementedError
-    def psi0(self,Z,posterior_variational):
+    def psi0(self,Z,variational_posterior):
         raise NotImplementedError
-    def psi1(self,Z,posterior_variational):
+    def psi1(self,Z,variational_posterior):
         raise NotImplementedError
-    def psi2(self,Z,posterior_variational):
+    def psi2(self,Z,variational_posterior):
         raise NotImplementedError
     def gradients_X(self, dL_dK, X, X2):
         raise NotImplementedError
@@ -49,16 +49,16 @@ class Kern(Parameterized):
         self._collect_gradient(target)
         self._set_gradient(target)
 
-    def update_gradients_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, posterior_variational):
+    def update_gradients_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
         """Set the gradients of all parameters when doing variational (M) inference with uncertain inputs."""
         raise NotImplementedError
     def gradients_Z_sparse(self, dL_dKmm, dL_dKnm, dL_dKdiag, X, Z):
         grad = self.gradients_X(dL_dKmm, Z)
         grad += self.gradients_X(dL_dKnm.T, Z, X)
         return grad
-    def gradients_Z_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, posterior_variational):
+    def gradients_Z_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
         raise NotImplementedError
-    def gradients_q_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, posterior_variational):
+    def gradients_q_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
         raise NotImplementedError
     
     def plot_ARD(self, *args, **kw):

GPy/kern/_src/linear.py

@@ -106,52 +106,52 @@ class Linear(Kern):
     #              variational              #
     #---------------------------------------#
 
-    def psi0(self, Z, posterior_variational):
+    def psi0(self, Z, variational_posterior):
-        return np.sum(self.variances * self._mu2S(posterior_variational), 1)
+        return np.sum(self.variances * self._mu2S(variational_posterior), 1)
 
-    def psi1(self, Z, posterior_variational):
+    def psi1(self, Z, variational_posterior):
-        return self.K(posterior_variational.mean, Z) #the variance, it does nothing
+        return self.K(variational_posterior.mean, Z) #the variance, it does nothing
 
-    def psi2(self, Z, posterior_variational):
+    def psi2(self, Z, variational_posterior):
         ZA = Z * self.variances
-        ZAinner = self._ZAinner(posterior_variational, Z)
+        ZAinner = self._ZAinner(variational_posterior, Z)
         return np.dot(ZAinner, ZA.T)
 
-    def update_gradients_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, posterior_variational, Z):
+    def update_gradients_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, variational_posterior, Z):
-        mu, S = posterior_variational.mean, posterior_variational.variance
+        mu, S = variational_posterior.mean, variational_posterior.variance
         # psi0:
-        tmp = dL_dpsi0[:, None] * self._mu2S(posterior_variational)
+        tmp = dL_dpsi0[:, None] * self._mu2S(variational_posterior)
         if self.ARD: grad = tmp.sum(0)
         else: grad = np.atleast_1d(tmp.sum())
         #psi1
         self.update_gradients_full(dL_dpsi1, mu, Z)
         grad += self.variances.gradient
         #psi2
-        tmp = dL_dpsi2[:, :, :, None] * (self._ZAinner(posterior_variational, Z)[:, :, None, :] * (2. * Z)[None, None, :, :])
+        tmp = dL_dpsi2[:, :, :, None] * (self._ZAinner(variational_posterior, Z)[:, :, None, :] * (2. * Z)[None, None, :, :])
         if self.ARD: grad += tmp.sum(0).sum(0).sum(0)
         else: grad += tmp.sum()
         #from Kmm
         self.update_gradients_full(dL_dKmm, Z, None)
         self.variances.gradient += grad
 
-    def gradients_Z_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, posterior_variational, Z):
+    def gradients_Z_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, variational_posterior, Z):
         # Kmm
         grad = self.gradients_X(dL_dKmm, Z, None)
         #psi1
-        grad += self.gradients_X(dL_dpsi1.T, Z, posterior_variational.mean)
+        grad += self.gradients_X(dL_dpsi1.T, Z, variational_posterior.mean)
         #psi2
-        self._weave_dpsi2_dZ(dL_dpsi2, Z, posterior_variational, grad)
+        self._weave_dpsi2_dZ(dL_dpsi2, Z, variational_posterior, grad)
         return grad
 
-    def gradients_q_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, posterior_variational, Z):
+    def gradients_q_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, variational_posterior, Z):
-        grad_mu, grad_S = np.zeros(posterior_variational.mean.shape), np.zeros(posterior_variational.mean.shape)
+        grad_mu, grad_S = np.zeros(variational_posterior.mean.shape), np.zeros(variational_posterior.mean.shape)
         # psi0
-        grad_mu += dL_dpsi0[:, None] * (2.0 * posterior_variational.mean * self.variances)
+        grad_mu += dL_dpsi0[:, None] * (2.0 * variational_posterior.mean * self.variances)
         grad_S += dL_dpsi0[:, None] * self.variances
         # psi1
         grad_mu += (dL_dpsi1[:, :, None] * (Z * self.variances)).sum(1)
         # psi2
-        self._weave_dpsi2_dmuS(dL_dpsi2, Z, posterior_variational, grad_mu, grad_S)
+        self._weave_dpsi2_dmuS(dL_dpsi2, Z, variational_posterior, grad_mu, grad_S)
 
         return grad_mu, grad_S
 

GPy/kern/_src/prod.py

@@ -42,10 +42,6 @@ class Prod(Kern):
         self.k1.update_gradients_full(dL_dK*self.k2(X[:,self.slice2]), X[:,self.slice1])
         self.k2.update_gradients_full(dL_dK*self.k1(X[:,self.slice1]), X[:,self.slice2])
 
-    def update_gradients_sparse(self, dL_dKmm, dL_dKnm, dL_dKdiag, X, Z):
-        self.k1.update_gradients_sparse(dL_dKmm * self.k2.K(Z[:,self.slice2]), dL_dKnm * self.k2(X[:,self.slice2], Z[:,self.slice2]), dL_dKdiag * self.k2.Kdiag(X[:,self.slice2]), X[:,self.slice1], Z[:,self.slice1] )
-        self.k2.update_gradients_sparse(dL_dKmm * self.k1.K(Z[:,self.slice1]), dL_dKnm * self.k1(X[:,self.slice1], Z[:,self.slice1]), dL_dKdiag * self.k1.Kdiag(X[:,self.slice1]), X[:,self.slice2], Z[:,self.slice2] )
-
     def gradients_X(self, dL_dK, X, X2=None):
         target = np.zeros(X.shape)
         if X2 is None:

GPy/kern/_src/rbf.py

@@ -40,27 +40,27 @@ class RBF(Stationary):
         self._Z, self._mu, self._S = np.empty(shape=(3, 1)) # cached versions of Z,mu,S
 
 
-    def psi0(self, Z, posterior_variational):
+    def psi0(self, Z, variational_posterior):
-        return self.Kdiag(posterior_variational.mean)
+        return self.Kdiag(variational_posterior.mean)
 
-    def psi1(self, Z, posterior_variational):
+    def psi1(self, Z, variational_posterior):
-        mu = posterior_variational.mean
+        mu = variational_posterior.mean
-        S = posterior_variational.variance
+        S = variational_posterior.variance
         self._psi_computations(Z, mu, S)
         return self._psi1
 
-    def psi2(self, Z, posterior_variational):
+    def psi2(self, Z, variational_posterior):
-        mu = posterior_variational.mean
+        mu = variational_posterior.mean
-        S = posterior_variational.variance
+        S = variational_posterior.variance
         self._psi_computations(Z, mu, S)
         return self._psi2
 
-    def update_gradients_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, posterior_variational):
+    def update_gradients_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
         #contributions from Kmm
         sself.update_gradients_full(dL_dKmm, Z)
 
-        mu = posterior_variational.mean
+        mu = variational_posterior.mean
-        S = posterior_variational.variance
+        S = variational_posterior.variance
         self._psi_computations(Z, mu, S)
         l2 = self.lengthscale **2
 
@@ -87,9 +87,9 @@ class RBF(Stationary):
         else:
             self.lengthscale.gradient += dpsi2_dlength.sum(0).sum(0).sum(0)
 
-    def gradients_Z_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, posterior_variational):
+    def gradients_Z_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
-        mu = posterior_variational.mean
+        mu = variational_posterior.mean
-        S = posterior_variational.variance
+        S = variational_posterior.variance
         self._psi_computations(Z, mu, S)
         l2 = self.lengthscale **2
 
@@ -108,9 +108,9 @@ class RBF(Stationary):
 
         return grad
 
-    def gradients_q_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, posterior_variational):
+    def gradients_q_variational(self, dL_dKmm, dL_dpsi0, dL_dpsi1, dL_dpsi2, Z, variational_posterior):
-        mu = posterior_variational.mean
+        mu = variational_posterior.mean
-        S = posterior_variational.variance
+        S = variational_posterior.variance
         self._psi_computations(Z, mu, S)
         l2 = self.lengthscale **2
         #psi1

GPy/kern/_src/static.py

@@ -43,7 +43,7 @@ class Static(Kern):
 
 class White(Static):
     def __init__(self, input_dim, variance=1., name='white'):
-        super(White, self).__init__(input_dim, name)
+        super(White, self).__init__(input_dim, variance, name)
 
     def K(self, X, X2=None):
         if X2 is None:
@@ -66,7 +66,7 @@ class White(Static):
 
 class Bias(Static):
     def __init__(self, input_dim, variance=1., name='bias'):
-        super(Bias, self).__init__(input_dim, name)
+        super(Bias, self).__init__(input_dim, variance, name)
 
     def K(self, X, X2=None):
         shape = (X.shape[0], X.shape[0] if X2 is None else X2.shape[0])

GPy/models/sparse_gp_regression.py

@@ -7,6 +7,7 @@ from ..core import SparseGP
 from .. import likelihoods
 from .. import kern
 from ..inference.latent_function_inference import VarDTC
+from ..util.misc import param_to_array
 
 class SparseGPRegression(SparseGP):
     """
@@ -33,18 +34,18 @@ class SparseGPRegression(SparseGP):
 
         # kern defaults to rbf (plus white for stability)
         if kernel is None:
-            kernel = kern.rbf(input_dim)#  + kern.white(input_dim, variance=1e-3)
+            kernel = kern.RBF(input_dim)#  + kern.white(input_dim, variance=1e-3)
 
         # Z defaults to a subset of the data
         if Z is None:
             i = np.random.permutation(num_data)[:min(num_inducing, num_data)]
-            Z = X[i].copy()
+            Z = param_to_array(X)[i].copy()
         else:
             assert Z.shape[1] == input_dim
 
         likelihood = likelihoods.Gaussian()
 
-        SparseGP.__init__(self, X, Y, Z, kernel, likelihood, X_variance=X_variance, inference_method=VarDTC())
+        SparseGP.__init__(self, X, Y, Z, kernel, likelihood, inference_method=VarDTC())
 
     def _getstate(self):
         return SparseGP._getstate(self)