add ss_mrd model

GPy/kern/_src/psi_comp/rbf_psi_gpucomp.py

@@ -259,6 +259,11 @@ class PSICOMP_RBF_GPU(PSICOMP_RBF):
         self.g_compDenom = module.get_function('compDenom')
         self.g_compDenom.prepare('PPPPii')
         
+    def __deepcopy__(self, memo):
+        s = PSICOMP_RBF_GPU(threadnum=self.threadnum, blocknum=self.blocknum, GPU_direct=self.GPU_direct)
+        memo[id(self)] = s 
+        return s
+    
     def _initGPUCache(self, N, M, Q):            
         if self.gpuCache == None:
             self.gpuCache = {

GPy/kern/_src/psi_comp/ssrbf_psi_gpucomp.py

@@ -309,6 +309,11 @@ class PSICOMP_SSRBF_GPU(PSICOMP_RBF):
         self.g_compDenom = module.get_function('compDenom')
         self.g_compDenom.prepare('PPPPPPPii')
 
+    def __deepcopy__(self, memo):
+        s = PSICOMP_SSRBF_GPU(threadnum=self.threadnum, blocknum=self.blocknum, GPU_direct=self.GPU_direct)
+        memo[id(self)] = s 
+        return s
+
     def _initGPUCache(self, N, M, Q):            
         if self.gpuCache == None:
             self.gpuCache = {

GPy/kern/_src/rbf.py

@@ -43,6 +43,15 @@ class RBF(Stationary):
     def __setstate__(self, state):
         return super(RBF, self).__setstate__(state)
 
+#     def copy(self):
+#         k = super(RBF, self).copy()
+#         # Make sure the copy of the kernel instance has different instance of psicomp
+#         if k.useGPU:
+#             k.psicomp = PSICOMP_RBF_GPU()
+#         else:
+#             k.psicomp = PSICOMP_RBF()
+#         return k
+
     #---------------------------------------#
     #             PSI statistics            #
     #---------------------------------------#

GPy/models/__init__.py

@@ -17,3 +17,4 @@ from ss_gplvm import SSGPLVM
 from gp_coregionalized_regression import GPCoregionalizedRegression
 from sparse_gp_coregionalized_regression import SparseGPCoregionalizedRegression
 from gp_heteroscedastic_regression import GPHeteroscedasticRegression
+from ss_mrd import SSMRD

GPy/models/ss_gplvm.py

@@ -24,7 +24,7 @@ class SSGPLVM(SparseGP):
 
     """
     def __init__(self, Y, input_dim, X=None, X_variance=None, init='PCA', num_inducing=10,
-                 Z=None, kernel=None, inference_method=None, likelihood=None, name='Spike-and-Slab GPLVM', group_spike=False, mpi_comm=None, **kwargs):
+                 Z=None, kernel=None, inference_method=None, likelihood=None, name='Spike_and_Slab GPLVM', group_spike=False, mpi_comm=None, **kwargs):
 
         self.mpi_comm = mpi_comm
         self.__IN_OPTIMIZATION__ = False

GPy/models/ss_mrd.py

@@ -0,0 +1,22 @@
+"""
+The Maniforld Relevance Determination model with the spike-and-slab prior
+"""
+
+from ..core import Model
+from .ss_gplvm import SSGPLVM
+
+class SSMRD(Model):
+    
+    def __init__(self, Ylist, input_dim, X=None, X_variance=None,
+                 initx = 'PCA', initz = 'permute',
+                 num_inducing=10, Z=None, kernel=None,
+                 inference_method=None, likelihoods=None, name='ss_mrd', Ynames=None):
+        super(SSMRD, self).__init__(name)
+        
+        self.updates = False
+        self.models = [SSGPLVM(y, input_dim, X=X, X_variance=X_variance, num_inducing=num_inducing,Z=Z,init=initx,
+                               kernel=kernel if kernel else None,inference_method=inference_method,likelihood=likelihoods,
+                               name='model_'+str(i)) for i,y in enumerate(Ylist)]
+        self.add_parameters(*(self.models))
+        self.updates = True
+

GPy/util/gpu_init.py

@@ -17,7 +17,7 @@ except:
     pass
 
 try:
-    if MPI_enabled: #and MPI.COMM_WORLD.size>1:
+    if MPI_enabled and MPI.COMM_WORLD.size>1:
         from .parallel import get_id_within_node
         gpuid = get_id_within_node()
         import pycuda.driver