some minor improvements in visualize

GPy/examples/dimensionality_reduction.py

@@ -60,7 +60,7 @@ def GPLVM_oil_100(optimize=True,M=15):
     m.plot_latent(labels=m.data_labels)
     return m
 
-def BGPLVM_oil(optimize=True,N=100,Q=10,M=15):
+def BGPLVM_oil(optimize=True,N=100,Q=10,M=15,max_f_eval=300):
     data = GPy.util.datasets.oil()
 
     # create simple GP model
@@ -72,10 +72,10 @@ def BGPLVM_oil(optimize=True,N=100,Q=10,M=15):
     if optimize:
         m.constrain_fixed('noise',0.05)
         m.ensure_default_constraints()
-        m.optimize('scg',messages=1)
+        m.optimize('scg',messages=1,max_f_eval=max(80,max_f_eval))
         m.unconstrain('noise')
         m.constrain_positive('noise')
-        m.optimize('scg',messages=1)
+        m.optimize('scg',messages=1,max_f_eval=max(0,max_f_eval-80))
     else:
         m.ensure_default_constraints()
 
@@ -120,7 +120,7 @@ def brendan_faces():
 def stick():
     data = GPy.util.datasets.stick()
     m = GPy.models.GPLVM(data['Y'], 2)
-    
+
     # optimize
     m.ensure_default_constraints()
     m.optimize(messages=1, max_f_eval=10000)

GPy/kern/rbf.py

@@ -173,7 +173,7 @@ class rbf(kernpart):
         """Think N,M,M,Q """
         self._psi_computations(Z,mu,S)
         tmp = self._psi2[:,:,:,None]/self.lengthscale2/self._psi2_denom
-        target_mu += (dL_dpsi2[:,:,:,None]*-tmp*2.*self._psi2_mudist).sum(1).sum(1)
+        target_mu += -2.*(dL_dpsi2[:,:,:,None]*tmp*self._psi2_mudist).sum(1).sum(1)
         target_S += (dL_dpsi2[:,:,:,None]*tmp*(2.*self._psi2_mudist_sq-1)).sum(1).sum(1)
 
 
@@ -207,7 +207,6 @@ class rbf(kernpart):
         if not (np.all(Z==self._Z) and np.all(mu==self._mu) and np.all(S==self._S)):
             #something's changed. recompute EVERYTHING
 
-            #TODO: make more efficient for large Q (using NDL's dot product trick)
             #psi1
             self._psi1_denom = S[:,None,:]/self.lengthscale2 + 1.
             self._psi1_dist = Z[None,:,:]-mu[:,None,:]
@@ -250,7 +249,7 @@ class rbf(kernpart):
         _psi2_denom = self._psi2_denom.squeeze()
         code = """
         double tmp;
-        
+
         #pragma omp parallel for private(tmp)
         for (int n=0; n<N; n++){
             for (int m=0; m<M; m++){
@@ -287,8 +286,8 @@ class rbf(kernpart):
         #include <omp.h>
         #include <math.h>
         """
-        weave.inline(code, support_code=support_code, libraries=['gomp'], 
+        weave.inline(code, support_code=support_code, libraries=['gomp'],
                      arg_names=['N','M','Q','mu','Zhat','mudist_sq','mudist','lengthscale2','_psi2_denom','psi2_Zdist_sq','psi2_exponent','half_log_psi2_denom','psi2','variance_sq'],
                      type_converters=weave.converters.blitz,**weave_options)
-        
+
         return mudist,mudist_sq, psi2_exponent, psi2

GPy/models/Bayesian_GPLVM.py

@@ -95,3 +95,4 @@ class Bayesian_GPLVM(sparse_GP, GPLVM):
             input_1, input_2 = which_indices
         ax = GPLVM.plot_latent(self, which_indices=[input_1, input_2],*args, **kwargs)
         ax.plot(self.Z[:, input_1], self.Z[:, input_2], '^w')
+        return ax

GPy/util/visualize.py

@@ -4,7 +4,7 @@ import GPy
 import numpy as np
 
 class lvm:
-    def __init__(self, model, data_visualize, latent_axis):
+    def __init__(self, model, data_visualize, latent_axis, latent_index=[0,1], latent_dim=2):
         self.cid = latent_axis.figure.canvas.mpl_connect('button_press_event', self.on_click)
         self.cid = latent_axis.figure.canvas.mpl_connect('motion_notify_event', self.on_move)
         self.data_visualize = data_visualize
@@ -12,6 +12,8 @@ class lvm:
         self.latent_axis = latent_axis
         self.called = False
         self.move_on = False
+        self.latent_index = latent_index
+        self.latent_dim = latent_dim
 
     def on_click(self, event):
         #print 'click', event.xdata, event.ydata
@@ -32,7 +34,8 @@ class lvm:
         if self.called and self.move_on:
             # Call modify code on move
             #print 'move', event.xdata, event.ydata
-            latent_values = np.array((event.xdata, event.ydata))
+            latent_values = np.zeros((1,self.latent_dim))
+            latent_values[0,self.latent_index] = np.array([event.xdata, event.ydata])
             y = self.model.predict(latent_values)[0]
             self.data_visualize.modify(y)
             #print 'y', y
@@ -45,7 +48,7 @@ class data_show:
         # If no axes are defined, create some.
         if axis==None:
             fig = plt.figure()
-            self.axis = fig.add_subplot(111)            
+            self.axis = fig.add_subplot(111)
         else:
             self.axis = axis
 
@@ -57,7 +60,7 @@ class vector_show(data_show):
     def __init__(self, vals, axis=None):
         data_show.__init__(self, vals, axis)
         self.vals = vals.T
-        self.handle = plt.plot(np.arange(0, len(vals))[:, None], self.vals)[0]
+        self.handle = self.axis.plot(np.arange(0, len(vals))[:, None], self.vals)[0]
 
     def modify(self, vals):
         xdata, ydata = self.handle.get_data()
@@ -84,7 +87,7 @@ class image_show(data_show):
         self.handle.set_array(self.vals)
         #self.axis.figure.canvas.draw()
         plt.show()
-        
+
     def set_image(self, vals):
         self.vals = np.reshape(vals, self.dimensions, order='F')
         if self.transpose:
@@ -94,7 +97,7 @@ class image_show(data_show):
         #if self.invert:
         #    self.vals = -self.vals
 
-class stick_show(data_show): 
+class stick_show(data_show):
     """Show a three dimensional point cloud as a figure. Connect elements of the figure together using the matrix connect."""
 
     def __init__(self, vals, axis=None, connect=None):
@@ -159,6 +162,6 @@ class stick_show(data_show):
             self.line_handle = self.axis.plot(np.array(x), np.array(y), np.array(z), 'b-')
 
         self.axis.figure.canvas.draw()
-        
+