Propagating temperature parameter

Author: alexbw

examples/library-mm.py

@@ -14,7 +14,7 @@
 
 T = 1000
 
-f = np.load('/Users/mattjj/Dropbox/Test Data/TMT_50p_mixtures_and_data.npz')
+f = np.load("/Users/Alex/Dropbox/Science/Datta lab/Posture Tracking/Test Data/TMT_50p_mixtures_and_data.npz")
 mus = f['mu']
 sigmas = f['sigma']
 data = f['data'][:T]
@@ -87,12 +87,12 @@
         # for the GMMs described above. roughly, gamma controls the total number
         # of states while alpha controls the diversity of the transition
         # distributions.
-        alpha=10.,gamma=10.,
+        # alpha=10.,gamma=10.,
         # NOTE: as with a_0 and b_0 for the GMMs described above, we can also
         # put gamma priors over alpha and gamma by commenting out the direct
         # alpha= and gamma= lines and using these instead
-        # alpha_a_0=1.,alpha_b_0=1./10,
-        # gamma_a_0=1.,gamma_b_0=1./10,
+        alpha_a_0=1.,alpha_b_0=1./10,
+        gamma_a_0=1.,gamma_b_0=1./10,
         obs_distns=hsmm_obs_distns,
         dur_distns=dur_distns)
 model.trans_distn.max_likelihood([rle(labels)[0]])

models.py

@@ -104,15 +104,15 @@ def _clear_caches(self):
 
     ### Gibbs sampling
 
-    def resample_model(self):
-        self.resample_obs_distns()
+    def resample_model(self, temp=None):
+        self.resample_obs_distns(temp=temp)
         self.resample_trans_distn()
         self.resample_init_state_distn()
         self.resample_states()
 
-    def resample_obs_distns(self):
+    def resample_obs_distns(self, temp=None):
         for state, distn in enumerate(self.obs_distns):
-            distn.resample([s.data[s.stateseq == state] for s in self.states_list])
+            distn.resample([s.data[s.stateseq == state] for s in self.states_list], temp=temp)
         self._clear_caches()
 
     def resample_trans_distn(self):
@@ -381,9 +381,9 @@ def generate(self,T,keep=True,**kwargs):
 
     ### Gibbs sampling
 
-    def resample_model(self):
+    def resample_model(self, temp=None):
         self.resample_dur_distns()
-        super(HSMM,self).resample_model()
+        super(HSMM,self).resample_model(temp=temp)
 
     def resample_dur_distns(self):
         for state, distn in enumerate(self.dur_distns):