Merge branch 'devel' into params

Conflicts:
	GPy/core/transformations.py
	GPy/kern/parts/kernpart.py

GPy/util/linalg.py

@@ -51,7 +51,7 @@ def dpotri(A, lower=0):
 
     :param A: Matrix A
     :param lower: is matrix lower (true) or upper (false)
-    :returns:
+    :returns: A inverse
     """
     return lapack.dpotri(A, lower=lower)
 

GPy/util/multioutput.py

@@ -0,0 +1,35 @@
+import numpy as np
+import warnings
+from .. import kern
+
+def build_lcm(input_dim, num_outputs, CK = [], NC = [], W_columns=1,W=None,kappa=None):
+    #TODO build_icm or build_lcm
+    """
+    Builds a kernel for a linear coregionalization model
+
+    :input_dim: Input dimensionality
+    :num_outputs: Number of outputs
+    :param CK: List of coregionalized kernels (i.e., this will be multiplied by a coregionalize kernel).
+    :param K: List of kernels that will be added up together with CK, but won't be multiplied by a coregionalize kernel
+    :param W_columns: number tuples of the corregionalization parameters 'coregion_W'
+    :type W_columns: integer
+    """
+
+    for k in CK:
+        if k.input_dim <> input_dim:
+            k.input_dim = input_dim
+            warnings.warn("kernel's input dimension overwritten to fit input_dim parameter.")
+
+    for k in NC:
+        if k.input_dim <> input_dim + 1:
+            k.input_dim = input_dim + 1
+            warnings.warn("kernel's input dimension overwritten to fit input_dim parameter.")
+
+    kernel = CK[0].prod(kern.coregionalize(num_outputs,W_columns,W,kappa),tensor=True)
+    for k in CK[1:]:
+        k_coreg = kern.coregionalize(num_outputs,W_columns,W,kappa)
+        kernel += k.prod(k_coreg,tensor=True)
+    for k in NC:
+        kernel += k
+
+    return kernel

GPy/util/univariate_Gaussian.py

@@ -32,4 +32,15 @@ def std_norm_cdf(x):
     x = float(x)
     return weave.inline(code,arg_names=['x'],support_code=support_code)
 
+def inv_std_norm_cdf(x):
+    """
+    Inverse cumulative standard Gaussian distribution
+    Based on Winitzki, S. (2008)
+    """
+    z = 2*x -1
+    ln1z2 = np.log(1-z**2)
+    a = 8*(np.pi -3)/(3*np.pi*(4-np.pi))
+    b = 2/(np.pi * a) + ln1z2/2
+    inv_erf = np.sign(z) * np.sqrt( np.sqrt(b**2 - ln1z2/a) - b )
+    return np.sqrt(2) * inv_erf